CN101401454A - Stereophonic sound imaging - Google Patents

Abstract

A method for reducing phase differences varying with frequency occurring at certain listening positions with respect to loudspeakers reproducing respective ones of multiple sound channels in a listening space, the phase differences occurring in a sequence of frequency bands in which the phase differences alternate between being predominantly in-phase and predominantly out-of-phase, comprises adjusting the phase in multiple frequency bands in which the multiple sound channels are out-of-phase at such listening positions. Such adjustment of phase includes the frequency bands in which the width of comb filtering pass bands and notches resulting from phase differences at such listening positions would be greater than or commensurate with the critical band width if the phase adjustment were not applied. The listening space may be the interior of a vehicle.

Description

Stereo imaging

Technical Field

The present invention relates to audio signal processing. More particularly, the present invention relates to improving the perceived image and sound image direction presented using a stereophonic playback system, particularly where the two listening positions are symmetrical with respect to a center line of such a stereophonic playback system. Aspects of the present invention include apparatus, methods and computer programs stored on computer readable media for causing a computer to perform the methods.

Background

Two-channel stereophonic playback systems are nearly ubiquitous in many environments, including live audio, home music playback, and car audio. The common effect is that the sound emitted by a pair of stereo speakers sounds differently at different listening positions relative to the speakers. These variations are mainly caused by the time difference it takes for the sound to reach the listening position from each loudspeaker and acoustically join up at the listening position. Secondary effects include the interaction of sound with the room, but these effects are not discussed here.

The time difference at the listening position is equivalent to a phase difference that varies with frequency. For the following discussion, the term "inter-speaker differential phase" (IDP) is defined as the phase difference of sound arriving at a listening position from a pair of stereo speakers.

Since the sound rendered by the two loudspeakers takes the same amount of time to reach the ears of a listener located equidistant from the two loudspeakers, the listener experiences substantially no IDP (see fig. 1 a). An IDP with a linearly increasing amplitude with frequency is experienced from a listener offset from a pair of stereo loudspeakers, i.e. closer to one of the loudspeakers (see fig. 2 a).

The variation of the IDP results in audible undesirable effects including comb filtering and blurring of the imaging of the audio signal presented by a pair of stereo speakers. A simple solution is to delay the signal presented through the closer loudspeakers. The amount of delay used is such that the signals presented through the two loudspeakers arrive at the listener's ears simultaneously. As a result, the IDP for that listener is zero and the listener does not experience undesirable imaging artifacts.

However, the use of simple delays is not suitable for environments such as vehicles where two listeners may be symmetrically off center with respect to a pair of stereo speakers-i.e., one listener is closer to the left speaker and the other listener is closer to the right speaker (see fig. 3). In such an environment, using a delay to correct the IDP of one listener may cause the experience of another listener to be degraded, as the rate of change of the IDP over frequency increases. The resulting effect may be unnatural enough to cause significant discomfort to other listeners.

For audio signals of directionality and imaging importance, i.e. signals with significant stationary components, an alternative to time correction is to adjust the IDP directly, i.e. to adjust the phase of the various frequencies. The phase is cyclic for each frequency. That is the mapping of the phase of an arbitrary value onto a 360 deg. circular space. For this decomposition, the phase values are limited to-180, and the range is 360 in total. To give a cyclical example, consider a phase value of 827 ° or 2 × 360+107 °, 827 ° or 2 × 360+107 ° being equivalent to 107 °. Similarly, -392 ° or-1X 360-32 ° is equivalent to-32 °. For reasons discussed below, frequencies closer to 0 ° (i.e., -90 °) than to-180 ° or 180 ° are considered to be "in phase" or enhanced, and frequencies closer to-180 ° or 180 ° than to 0 ° (i.e., 90 ° -180 ° or 90 ° -180 °) are considered to be "out of phase" or cancelled (see fig. 4a and 4 b).

In a typical vehicular environment, the IDP of each listener is as follows. Frequencies between 0Hz and about 250Hz are predominantly in phase-i.e., the IDP is between-90 and 90. Frequencies between about 250Hz and 750Hz are mainly out of phase-i.e., IDP is between 90 and 180 or-90 and-180. Frequencies between about 750Hz and 1250Hz are mainly in phase. This alternating sequence of mainly in-phase and mainly out-of-phase continues with increasing frequency up to the limit of the human ear of about 20 kHz. In this example, the cycle repeats every 1 kHz. The exact band start and end frequencies are a function of the interior dimensions of the vehicle and the location of the listener.

It is generally accepted that the human auditory system is sensitive to phase differences up to about 1500 Hz. Thus, below about 1500Hz, the change in IDP results in a significant distortion of the apparent spatial direction or image of the audio signal. This is in addition to the amplitude distortion caused by comb filtering, which is audible both below and above 1500 Hz.

It is also commonly understood that the human auditory system breaks down the broad spectrum into a number of smaller groups of frequencies or bands, referred to as critical bands. The critical band represents the minimum frequency difference at which the two frequencies can still be easily heard apart, which varies with frequency. At low frequencies, the critical band is very narrow and widens as the frequency increases. In the following discussion, "band" refers to a frequency band in which sounds arriving at a listener from multiple speakers are in phase and out of phase. In the following discussion, the critical zone is referred to as the "critical zone".

In the above-described vehicle environment, since the widths of the peaks and valleys at about 500Hz are equal to or greater than the critical band width, the comb filtering effect can be clearly heard for frequencies below about 4 kHz. Above about 6kHz, the critical bandwidth becomes larger than the combined width of one peak and one valley, and the comb filtering effect becomes substantially inaudible.

Thus, according to an aspect of the present invention, it is preferred that the IDP is adjusted for frequencies up to a critical bandwidth that becomes larger than the combined bandwidth of one peak and one valley of the comb filter, i.e. frequencies below about 6 kHz. This can be achieved by performing phase adjustment on a plurality of frequency bands in two channels of the audio signal so as to correct the inter-speaker differential phase at each listening position. Once applied, the resulting IDP observed in the listening position in the ideal case is within +/-90 ° for both listeners (see fig. 11a and 11 b). Reducing the IDP in this manner significantly improves perceptual imaging and reduces amplitude distortion from fully audible comb filtering with deep and wide nulls to relatively benign pulsations of +/-3dB, which are substantially inaudible to most listeners and sound content.

Many of the prior art methods only focus on IDPs below about 1 kHz. They try to correct the IDP for both listeners in the lowest frequency band where the sounds arriving at the listeners are mainly out of phase. They do this by adding substantially 180 to the IDP in this band using filters and phase shifters. As a result, the IDP for both listeners after correction was-90 to 90 ° below 1 kHz. That is, frequencies below 1kHz are primarily in phase for each listener, and the listener experiences greatly improved imaging. The main drawback of such methods is that they ignore IDP at higher frequencies where phase correction may be advantageous.

Us patent 4,817,162 teaches the use of filters and phase shifters in the two channels to add 180 ° to the relative phase of the signal between the left and right channels for frequencies in the range 200Hz to 600 Hz. In this teaching, this frequency range represents the first band where the sound arriving at the listener is mainly out of phase at the two listening positions (see fig. 5a and 5 b). A problem with this teaching is that the phase shifter does not provide a sufficiently fast rate of phase change at the band edge to provide substantial correction of IDP.

Us patent No. 5,033,092 teaches the use of filters and phase shifters to advance the phase of one channel by 60 to 90 and the phase of another channel by-60 to-90 in the frequency range of 200Hz to 1 kHz. In this teaching, 200Hz approximately represents the onset of the first band where the sound reaching the listener is predominantly out of phase. When each channel is advanced by 90 and-90 in this band, respectively, the total relative phase difference in this band is 180. The expected results are similar to the method of U.S. patent No. 4,817,162. An important advantage of this teaching is that the amplitude distortion in each channel is limited to a maximum of 3dB, since the phase of each channel is adjusted by 90 ° at the most. Conversely, if a phase shift of 180 ° is produced by filtering only one channel, then this channel will have an audible zero value in its magnitude response. That is, the magnitude response will drop to zero in the transition from 0 deg. to 180 deg., and vice versa.

Us patent 6,038,323 teaches the use of filters and phase shifters to add 180 ° to the phase of all frequencies above 300 Hz. In this teaching, 300Hz represents the beginning of the first band where the sound arriving at the listener is predominantly out of phase for each listening position. The reason for this teaching is that humans are not sensitive to IDP for frequencies above this first out-phasing band (see fig. 6a and 6b) in order to simplify the filter design, keeping frequencies above this first band out-of-phase. This teaching ignores the fact that for frequencies above this first band, the amplitude distortion caused by comb filtering can be heard.

Disclosure of Invention

It is an object of the invention to improve the perceptual imaging of audio signals presented by a stereo playback system to a listener positioned symmetrically off-center of the playback system. This is achieved by performing phase adjustment on a plurality of frequency bands in two channels of an audio signal, thereby correcting the inter-speaker differential phase at each listening position.

Drawings

Fig. 1a schematically shows a listening position and a spatial relationship of two loudspeakers in which the listening position is equidistant from the loudspeakers.

Fig. 1b shows the ideal interaural phase difference (IDP) for all frequencies at the equidistant listening positions of fig. 1 a. This example shows how the IDP at such listening position does not vary with frequency.

Fig. 2a schematically shows the spatial relationship of the listening position with respect to the offset of the two loudspeakers.

Fig. 2b shows the ideal interaural phase difference (IDP) for all frequencies at the listening position of fig. 2 a. This example shows how the IDP at the listening position varies with frequency.

Fig. 3 schematically shows the spatial relationship of two listening positions, each listening position being symmetrically offset with respect to two loudspeakers.

Fig. 4a and 4b show how the IDP varies with frequency for each of the two listening positions of fig. 3.

Fig. 5a and 5b show ideal IDP responses at two listening positions in a system implementing the teachings of U.S. patent No. 4,817,162.

Fig. 6a and 6b show the ideal IDP response at two listening positions in a system implementing the teachings of U.S. patent No. 6,038,323.

Fig. 7a shows a functional schematic block diagram of a possible FIR-based implementation of aspects of the invention applied to one of the two channels, in this case the left channel.

Fig. 7b shows a functional schematic block diagram of a possible FIR-based implementation of aspects of the invention applied to one of the two channels, in this case the right channel.

Fig. 8a is an idealized magnitude response of the signal output 703 of the filter or filter function 702 of fig. 7 a.

Fig. 8b is an idealized magnitude response of the signal output 709 of the subtractor or subtractor function 708 of fig. 7 a.

Fig. 9a is an idealized phase response of output signal 715 of fig. 7 a.

Fig. 9b is an idealized phase response of output signal 735 of fig. 7 b.

Fig. 9c is an idealized phase response representing the relative phase difference between the two output signals 715 (fig. 7a) and 735 (fig. 7 b).

Fig. 10a shows the tolerance of an ideal IDP compensation filter, which indicates its desired phase requirement.

Fig. 10b is the desired phase response used as input to the signature filter design algorithm.

Fig. 10c is a weighting function for the feature filter design algorithm.

Fig. 11a is an ideal IDP phase response for the left listening position of fig. 3 when the FIR filter of fig. 7a is employed.

Fig. 11b is the ideal IDP phase response for the right listening position of fig. 3 when the FIR filter of fig. 7b is employed.

Fig. 12 shows the magnitude response and the ideal phase response of an implementation of the FIR filter before optimization.

Fig. 13 shows the magnitude response and the ideal phase response of an implementation of an optimized FIR filter.

Fig. 14 shows the magnitude and phase response of an implementation of an IIR filter designed using the group delay method.

Fig. 15, 16 and 17 show the phase response for an implementation of the eigenfilter design algorithm with different values of h.

Fig. 18 is a schematic diagram showing an example of an all-pass filter lattice structure implementation.

Fig. 19 schematically shows the listening position and speaker layout of the front seat of the vehicle when the left, center and right speakers are present.

FIG. 20 schematically shows a functional block diagram of an aspect of the present invention as applied to the configuration of FIG. 19.

Fig. 21a schematically illustrates a four-channel speaker configuration having two listening positions in which aspects of the present invention may be employed.

Fig. 21b schematically shows a four-channel speaker configuration with four listening positions in which aspects of the present invention may be employed.

Fig. 21c schematically illustrates a six-channel speaker configuration having four listening positions in which aspects of the present invention may be employed.

Fig. 22a and 22b are functional block diagrams of a general filter bank implementation of the ideal filter whose tolerances are shown in fig. 10 a.

Fig. 23 shows the poles and zeros of an implementation of an IIR filter designed using the group delay method.

Fig. 24 and 25 show the implemented poles and zeros before and after the filter order reduction for an IIR filter designed using a characteristic filter design algorithm.

Fig. 26 shows the original desired phase response for the signature filter design algorithm.

Fig. 27 and 28 show the phase response of the implementation of an IIR filter designed using a characteristic filter design algorithm before and after the reduction of the filter order.

Fig. 29 shows the expected phase response of the pre-twist after five correction iterations.

Fig. 30 shows the phase response of an implementation of an IIR filter designed using a characteristic filter design algorithm after order reduction and five correction iterations.

Detailed Description

Fig. 1a shows the listening position and the spatial relationship of two loudspeakers. Listening position and left loudspeaker d₁The distance and listening position between and the right loudspeaker d₂The distances between them are equal. Lines representing other equidistant listening positions are also shown. Fig. 1b shows the interaural phase difference (IDP) for all frequencies at equidistant listening positions. In such equidistant locations, the perceived direction and imaging of the content presented through the speakers tends to be natural, as intended by the content creator.

Fig. 2a shows the spatial relationship of the listening position with respect to the offset of the two loudspeakers. In this example, the distance d between the listening position and the left loudspeaker₃Less than the distance d between the listening position and the right loudspeaker₄. Fig. 2b shows how the IDP at the listening position varies with frequency. Even if the IDP monotonically decreases, this figure (and all other IDP figures) shows equivalent values in the range of-180 deg.. At 0Hz, the signals are in phase and move out of phase with increasing frequency before returning to frequency a in phase. This phase period repeats with increasing frequency. The frequency a at which the repetition of the period occurs is directly related to the listening position and the difference in distance between the two loudspeakers. For example, if d is away from the left speaker₃Is 0.75 m and is d from the right loudspeaker₄Is 1.075 meters, the difference in distance is 0.325 meters. Frequency point a is equal to the speed of sound divided by the distance difference, or approximately 330 meters per second divided by 0.325, which yields 1015 Hz. Thus, in this example, the IDP period repeats every 1015 Hz.

Fig. 3 shows the spatial relationship of two listening positions, each listening position being symmetrically offset with respect to two loudspeakers. Fig. 4a and 4b show how the IDP varies with frequency for each of two listening positions. It can be seen that for each period of the IDP there are frequencies that are predominantly in phase and frequencies that are predominantly out of phase. That is, the IDP has a frequency of-90 to-90 DEG, and the IDP has a frequency of-90 to-180 DEG or 90 to 180 deg. IDP is mainly that the out of phase frequencies cause undesired audible effects including blurring of the imaging of the audio signal presented through the two loudspeakers.

Fig. 5a and 5b show an idealized representation of the effect of the teachings described in U.S. patent No. 4,817,162. This teaching adds 180 ° to the IDP of all frequencies in the first band, which are mainly out of phase. In this teaching, this band ranges from approximately 200Hz to 600 Hz. As can be seen in fig. 5a and 5b, these sounds are now predominantly in phase for both listening positions. However, this teaching ignores frequencies above 600Hz that are mainly out of phase. The teachings described in U.S. patent No. 5,033,092 are similar to those of U.S. patent No. 4,817,162, except that the frequency range treated is approximately 200Hz to 1 kHz.

Fig. 6a and 6b show an idealized representation of the effect of the teachings described in U.S. patent No. 6,038,323. This teaching adds 180 ° to the IDP of all frequencies in and above the first vocal cord that are primarily out of phase. In this teaching, this band starts at about 200 Hz. It can be seen in fig. 6a and 6b that the sound in this first band is now mainly in phase. However, this teaching also ignores that the higher frequency bands, which are mainly out of phase, reverse the position of the bands in phase and the bands in phase.

According to one aspect of the invention, the comb filtering effect heard at a particular listening position is minimized by correcting the IDPs of multiple frequency bands, which are mainly out of phase. Although the previous invention focused on the lowest out-of-phase band, significant hearing improvement can be achieved by correcting for IDP of multiple bands below frequencies approximately up to where the widths of the comb filter passbands and valleys approach the critical bandwidth. Above this frequency, no auditory improvement in imaging can be achieved by correcting the out-of-phase band. In a vehicle, this frequency is approximately 6kHz, but varies slightly with the actual interior dimensions of the vehicle and the relative distance to the speakers.

According to aspects of the present invention, the audio signal is divided into in-phase and out-of-phase frequency bands, and for each out-of-phase band, a 180 phase shift is added to the relative phase between the two channels. This is preferably done by shifting the phase by 90 ° in one channel and by-90 ° in the other channel. An alternative is to add 180 ° to the band in only one channel; however, this may cause significant, undesirable fluctuations in the amplitude response of the vocal tract.

Examples

In various exemplary embodiments of the present invention, a bank of filters provides a substantially flat magnitude response and a phase response that produces a combined phase shift between the channels with alternating bands of 0 ° and 180 °. To avoid undesired fluctuations in the magnitude response, the left channel may be given a 90 ° phase shift and the right channel a-90 ° phase shift (see fig. 9a, 9b, and 9 c). If this is done with a 180 phase transition in one channel, the amplitude will drop towards- ∞ dB in the phase transition. However, by using only a 90 ° transition, the maximum steep drop in frequency (dip) is about-3 dB. Above about 6kHz, the phase response is no longer as important and can be set to zero for both channels.

For some filter designs, particularly digital filter designs, it may be more efficient not to terminate the phase shift of the band at a defined frequency, but to continue phase shifting the band up to the nyquist frequency. For other designs, it may be more efficient to move the phase of only the minimum number of bands required to produce the desired result. For some implementations, the number of bands that are phase shifted may have little or no effect on efficiency, and the choice as to the number of bands that are phase shifted may be determined based on the total filter order and the resulting time smearing.

Based on the geometric positions depicted in fig. 1a, 2a and 3, the desired filter response is the frequency f_dFunction of, frequency f_dCorresponding to a wavelength equal to a path difference between the left speaker and the right speaker at the listening position off-center. This is shown in equation 1:

$f_d=\frac{c}{|d_L-d_R|}$

wherein d is_LDistance from listener to left speaker, d_RIs the distance from the listener to the right speaker, and c is the sound velocity (all distances are in meters).

The phase performance of the IDP compensation filter can be characterized by the tolerance in fig. 10a, where f is depicted in fig. 10a_dIs a frequency corresponding to a wavelength equal to the path difference; b is the number of bands; Δ F_beg、ΔF_midAnd Δ F_endThe transition widths before the first band, between all bands and after the last band, respectively; delta P_bndPhase error inside the band; delta P_beg、ΔP_midAnd Δ P_endPhase errors before the first band, between all bands and after the last band, respectively.

Although these tolerances are specified to be substantially equal across all bands, they may alternatively be specified differently for each band. For example, it may be advantageous to have very fast transitions for the first band, where the human ear is most sensitive to phase, with wider transitions as the frequency rises, to reduce the filter order and improve efficiency.

In summary, the filters may be implemented using a filter bank that divides the left and right audio signals into subbands in which alternate subbands are phase adjusted such that the relative phase in these subbands between the two channels is 180 °. Fig. 26a and 26b show examples of general filter bank implementations. Subbands that are not phase shifted may require delay processing to match their delay to any delay imparted by the phase shifting processing. The recombination of the subbands can be achieved by adding the subbands (see fig. 6a and 6b) or by an inverse filter bank.

Alternatively, the filter can be designed directly to give the desired phase response.

The following discussion of Finite Impulse Response (FIR) filters is followed by an example of filter bank based design; however, the filter bank method may use an Infinite Impulse Response (IIR) filter. Following the FIR filter discussion, many straightforward design approaches that can lead to a very efficient IIR filter are discussed.

Finite impulse response filter

IDP phase compensation for arrangements such as in the example of fig. 3 may be implemented using Finite Impulse Response (FIR) filters and linear phase digital filters or filter functions. Such filters or filter functions can be designed to achieve very predictable controlled phase and amplitude responses. Fig. 7a and 7b show block diagrams of possible FIR-based implementations of aspects of the invention as applied to one of the two channels, respectively.

In the example of fig. 7a, which processes the left channel, two complementary comb filtered signals are generated (at 703 and 709) which, if combined, will have a substantially flat magnitude response. Fig. 8a shows the comb filter response of a band pass filter or filter function ("BP filter") 702. One or more filters or filter functions may be used to obtain such a response. Fig. 8b shows the effective comb filter response resulting from the scheme of BP filter 702, time delay or delay function ("delay") 704 and subtractive combiner 708. In order for the comb filter responses to be substantially complementary, the BP filter 702 and the delay 704 should have substantially the same delay characteristics (see fig. 8a and 8 b). One of the comb filtered signals is phase shifted by 90 ° to give the desired phase adjustment in the desired frequency band. Although either of the two comb filtered signals may be shifted by 90 °, in this example the signal at 709 is phase shifted. The selection of one or the other of the motion signals affects the selection in the correlation process shown in the example of fig. 7b, so that the total shift from channel to channel is as desired. The use of a linear phase FIR filter allows the economical generation of two comb filtered signals (703 and 709) using a filter or filters selected only for a set of frequency bands as in the example of fig. 8 a. Preferably, the delay through the BP filter 702 does not change with frequency. This allows the complementary signals to be generated by: the original signal is delayed by the same amount of time as the group delay of the FIR BP filter 702 and the filtered signal is subtracted from the delayed original signal (in a subtraction combiner 708 as shown in fig. 7 a). Any frequency-invariant delays given by the 90 ° phase shifting process should be applied to the non-phase-adjusted signals before being combined to ensure a flat response again.

Filtered signal 709 is passed through a wideband 90 phase shifter or phase shifting process ("90 phase shift") 710 to produce signal 711. Signal 703 is delayed by a delay or delay function 712 having the same delay characteristics as 90 ° phase shift 710 to generate signal 713. In a summing summer or summing function 714, the 90 ° phase shifted signal 711 and the delayed signal 713 produce an output signal 715. The 90 ° phase shift may be implemented using any of a number of known methods, such as a hilbert transform. The output signal 715 has a substantially uniform gain with only a very narrow-3 dB dip between the unmodified and phase-shifted bands at frequencies corresponding to the transition points, but the output signal 715 has a phase response with frequency changes, as shown in fig. 9 a.

Fig. 7b shows a block diagram of aspects of the invention applied to the other of the two channels, in this case the right channel. This block diagram is very similar to that of the left channel except that the delayed signal (in this case signal 727) is subtracted from the filtered signal (in this case signal 723) and not vice versa. As shown in fig. 9b, the final output signal 735 has substantially uniform gain, but a-90 ° phase shift for the frequency band being phase shifted (compared to +90 ° in the left channel as shown in fig. 9 a).

The relative phase difference between the two output signals 715 and 715 is shown in fig. 9 c. The phase difference shows a 180 ° combined phase shift for each frequency band that is predominantly out of phase at each listening position. Thus, the out-of-phase bands become predominantly in-phase at the listening position. The resulting corrected IDP for each listening position (see fig. 3) is shown in fig. 11a and 11 b.

FIR amplitude and phase response

Due to the nature of FIR filters, it is not possible to produce an all-pass FIR filter (except for a pure delay). Thus, there is inevitably some deviation in the filter magnitude response. For the FIR implementation described above, fig. 12 and 13 provide magnitude and phase response examples for two different filter orders.

There is a-3 dB dip in the amplitude response during the transition region between bands. As the filter order increases, the width of the dip becomes smaller and the phase transition from +/-90 to 0 becomes faster. However, a larger filter order means a larger impulse response.

While FIR filters are easy to design, they have certain characteristics that are undesirable for implementing aspects of the present invention. First, they require a relatively long impulse response to achieve the required amplitude and phase response-long impulse responses result in high computational complexity. Second, long impulse responses lead to an undesirable temporal tail-off effect for the audible perception of an impulsive or impulsive audio signal.

Consideration of FIR implementation

For efficiency, the filters or filtering processes 702 and 722 in fig. 7a and 7b, respectively, are constructed as equally spaced comb filter banks followed by low pass filters. The comb filter can be efficiently implemented as a sparse FIR filter. A low pass filter may be employed to stop the phase adjustment of the band above the desired cut-off frequency.

Devices or processes 710 and 730 are 90 ° phase shift filters or filtering processes. For a filter that is suitable for most audio frequencies at sample rates of 44.1kHz and 48kHz, 400-800 filter taps are required. Since implementations using direct convolution are expensive, a Fast Fourier Transform (FFT) can be used to employ fast convolution.

Furthermore, for sample rates of 44.1kHz and 48kHz, the low pass filter of the filtering process should have 200-400 taps. It may also benefit from fast convolution and may be combined with a 90 ° phase shift filter or filter process.

Infinite impulse response filter

Preferred embodiments use an Infinite Impulse Response (IIR) all-pass filter to achieve the desired phase response. IIR filters have the advantage that they typically have a shorter impulse response than similar FIR filters for the desired phase and amplitude response. Shorter impulse responses result in reduced computational complexity and reduced time-smearing effects. However, IIR filters are difficult to design.

Group delay method

Most classical IIR filter design techniques focus on matching a particular magnitude response. However, there are several techniques for designing all-pass IIR filters. One approach for all-pass filter design is based on finding the smallest pth order that fits the desired group delay. Such a method may be implemented, for example, by using a computer tool such as MATLAB (a trademark of The Math Works, inc.). A MATLAB function iirgrapdelay.m, which is part of the filter design tool box, can be used. In implementing aspects of the invention, the ideal phase response is an alternating band with sharp transitions. Since the group delay is the first derivative of the phase, the ideal group delay is 0 in-band, and ± ∞ at the transition. Since such discontinuities are unlikely to fit into the minimum p-th order algorithm, an approximation of the ideal phase response with no discontinuities in the derivative must be found. By selecting the desired phase response as a sinusoid that is optimally aligned with the desired band, an IIR filter can be designed that approximates the desired response. Fig. 14 shows the magnitude and phase response of a filter designed using the group delay method.

However, the group delay algorithm becomes numerically unstable at larger orders and generally does not converge. Furthermore, since the algorithm is adapted to the group delay, any error in the group delay causes a larger error in the phase response due to the integration. Thus, there is a large number of trial and error methods or parameter searches to find a filter with the desired performance. In addition, since the method can only be designed for a small number of orders, the method may not be suitable for applications that require a large number of in-band phase adjustments. That is, a case where a distance difference to two speakers, i.e., a Δ distance, is large.

Characteristic filter method

Another technique for designing IIR all-pass filters is the eigenfilter approach. See, for example, the following technical papers: nguyen et al, "origin filter application for The Design of Allpass Filters application a Given Phase Response", IEEE Trans On Signal Processing, vol.42(9), 09/1994 and Tkacenko et al, "On The origin filter Method and Applications: a Tutorial ", IEEE Transactions on Circuits And Systems-II: analog And digital Signal Processing, Vol, 50, No.9, September1994,http://www.systems.caltech.edu/EE/Groups /dsp/students/andre/papers/journal/eigen tutorial.pdf。

the eigenfilter approach allows an approximate least squares fit to the desired phase response. Although it is not guaranteed that a stable filter is generated, it reliably produces a stable filter if the conditions are set appropriately. In addition, there are some iterative methods that are closer to true least squares or closer to phase equivalent fluctuations. The eigenfilter method is a very efficient technique since it can be numerically stable even up to large filter orders.

The eigenfilter method is based on finding an error metric that can be expressed as a quadratic form of the filter coefficients, such as e ═ a^TPa, where ε is the error and a is the denominator filter systemVector of numbers, P is a matrix. Once expressed as a formula, a can be found using Rayleigh's principle. Thus, the eigenvalues of P are proportional to the error epsilon and the eigenvector associated with the smallest eigenvalue is the best solution for a.

For all-pass filters, the total phase phi of the filter of order N_H(ω) phase φ from denominator by the following expression_A(ω) correlation:

φ_H(ω)＝-Nω-2φ_A(ω) (2)

where ω represents frequency in radians. An approximate estimate of the least-squares phase error for an all-pass filter is:

<math> <mrow> <mi>&epsiv;</mi> <mo>&ap;</mo> <mfrac> <mn>1</mn> <mi>&pi;</mi> </mfrac> <mo>&Integral;</mo> <mi>W</mi> <mrow> <mo>(</mo> <mi>&omega;</mi> <mo>)</mo> </mrow> <msup> <mrow> <mo>(</mo> <msup> <mi>a</mi> <mi>T</mi> </msup> <mi>s</mi> <mrow> <mo>(</mo> <mi>w</mi> <mo>)</mo> </mrow> <mo>)</mo> </mrow> <mn>2</mn> </msup> <mi>d&omega;</mi> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>3</mn> <mo>)</mo> </mrow> </mrow></math>

wherein,

s(ω)＝[sin(φ_A，des(ω))sin(φ_A，des(ω)+ω)...sin(φ_A，des(ω)+Nω)]^T (4)

w (ω) is the weight, φ, provided to the user_A，des(ω) is the desired phase of the denominator. From (1), there are

<math> <mrow> <msub> <mi>&phi;</mi> <mrow> <mi>A</mi> <mo>,</mo> <mi>des</mi> </mrow> </msub> <mrow> <mo>(</mo> <mi>&omega;</mi> <mo>)</mo> </mrow> <mo>=</mo> <mo>-</mo> <mfrac> <mn>1</mn> <mn>2</mn> </mfrac> <mrow> <mo>(</mo> <msub> <mi>&phi;</mi> <mrow> <mi>H</mi> <mo>,</mo> <mi>des</mi> </mrow> </msub> <mrow> <mo>(</mo> <mi>&omega;</mi> <mo>)</mo> </mrow> <mo>+</mo> <mi>N&omega;</mi> <mo>)</mo> </mrow> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>5</mn> <mo>)</mo> </mrow> </mrow></math>

Next, we can express the error metric, epsilon, as a quadratic expression:

ε＝a^Tpa, wherein, <math> <mrow> <mi>P</mi> <mo>=</mo> <mfrac> <mn>1</mn> <mi>&pi;</mi> </mfrac> <mo>&Integral;</mo> <mi>W</mi> <mrow> <mo>(</mo> <mi>&omega;</mi> <mo>)</mo> </mrow> <mi>s</mi> <mrow> <mo>(</mo> <mi>&omega;</mi> <mo>)</mo> </mrow> <msup> <mi>s</mi> <mi>T</mi> </msup> <mrow> <mo>(</mo> <mi>&omega;</mi> <mo>)</mo> </mrow> <mi>d&omega;</mi> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>6</mn> <mo>)</mo> </mrow> </mrow></math>

the integral can be approximated by a discrete sum:

<math> <mrow> <mi>P</mi> <mo>=</mo> <mfrac> <mn>1</mn> <mi>&pi;</mi> </mfrac> <munderover> <mi>&Sigma;</mi> <mrow> <mi>i</mi> <mo>=</mo> <mn>0</mn> </mrow> <mi>M</mi> </munderover> <mi>W</mi> <mrow> <mo>(</mo> <mfrac> <mi>i</mi> <mi>M</mi> </mfrac> <mi>&pi;</mi> <mo>)</mo> </mrow> <mi>s</mi> <mrow> <mo>(</mo> <mfrac> <mi>i</mi> <mi>M</mi> </mfrac> <mi>&pi;</mi> <mo>)</mo> </mrow> <msup> <mi>s</mi> <mi>T</mi> </msup> <mrow> <mo>(</mo> <mfrac> <mi>i</mi> <mi>M</mi> </mfrac> <mi>&pi;</mi> <mo>)</mo> </mrow> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>7</mn> <mo>)</mo> </mrow> </mrow></math>

wherein M is the division [0, π]The number of frequency steps. If λ_minIs the minimum eigenvalue of P, a_minFor the corresponding feature vector, the desired filter is then:

<math> <mrow> <mi>H</mi> <mrow> <mo>(</mo> <mi>z</mi> <mo>)</mo> </mrow> <mo>=</mo> <mfrac> <mrow> <munderover> <mi>&Sigma;</mi> <mrow> <mi>n</mi> <mo>=</mo> <mn>0</mn> </mrow> <mi>N</mi> </munderover> <msub> <mi>a</mi> <mi>min</mi> </msub> <mrow> <mo>[</mo> <mi>N</mi> <mo>-</mo> <mi>n</mi> <mo>]</mo> </mrow> <msup> <mi>z</mi> <mrow> <mo>-</mo> <mi>n</mi> </mrow> </msup> </mrow> <mrow> <munderover> <mi>&Sigma;</mi> <mrow> <mi>n</mi> <mo>=</mo> <mn>0</mn> </mrow> <mi>N</mi> </munderover> <msub> <mi>a</mi> <mi>min</mi> </msub> <mrow> <mo>[</mo> <mi>n</mi> <mo>]</mo> </mrow> <msup> <mi>z</mi> <mrow> <mo>-</mo> <mi>n</mi> </mrow> </msup> </mrow> </mfrac> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>8</mn> <mo>)</mo> </mrow> </mrow></math>

unfortunately, the resulting filter is not guaranteed to be stable. However, a stable filter can usually be found if the following constraints are applied:

φ_H，des(π)＝-Nπ (9)

characteristic filter method filter design

Based on the parameterizations given in fig. 10b and 10c, the following equations can be established to produce a filter that achieves the desired magnitude and phase response to provide IDP correction at the listening position.

The desired phase response of the left and right channels is given by the following equation:

the least squares weight is given by the following equation:

the number of bands B to be phase corrected is given by the following equation:

n is the number of sampling periods corresponding to the relative time delay:

$n=\frac{|d_L-d_R|}{c}{f}_s---\left(14\right)$

wherein f is_cA cut-off frequency above which no band is phase adjusted; f. of_dIs a frequency corresponding to a wavelength equal to the path difference; Δ f_beg、Δf_midAnd Δ f_endThe transition widths before the first band, between all bands and after the last band, respectively; omega_pre、ω_in、ω_outAnd ω_postUser-defined weights for before the first band, inside the band, between the bands, and after the last band, respectively; d_LAnd d_RIs the distance (in meters) from the listening position to both loudspeakers; c is the speed of sound (in m/s), f_sIs the sampling rate (in Hz).

For the left filter, there is a-pi/2 or-90 offset from the linear delay in the specified band, and the right filter has a + pi/2 or +90 offset. Can also confirm that phi_H，L，desAnd phi_H，R，desThis satisfies (9), which allows a stable filter to be reliably found. By choosing different weights, the transition width and filter order, the amount of fluctuation and the sharpness of the transition can be controlled.

Feature filter improvements

As described in the paper by t.q.nguyen et al, a closer approximation to the true least squares error can be obtained by using an iterative weighting function. This results in the following error metrics:

ε＝a_q ^TPa_qwherein <math> <mrow> <mi>P</mi> <mo>=</mo> <mfrac> <mn>1</mn> <mi>&pi;</mi> </mfrac> <mo>&Integral;</mo> <mi>W</mi> <mrow> <mo>(</mo> <mi>&omega;</mi> <mo>)</mo> </mrow> <mfrac> <mrow> <mi>s</mi> <mrow> <mo>(</mo> <mi>&omega;</mi> <mo>)</mo> </mrow> <msup> <mi>s</mi> <mi>T</mi> </msup> <mrow> <mo>(</mo> <mi>&omega;</mi> <mo>)</mo> </mrow> </mrow> <mrow> <msup> <msub> <mi>a</mi> <mrow> <mi>q</mi> <mo>-</mo> <mn>1</mn> </mrow> </msub> <mi>T</mi> </msup> <mi>c</mi> <mrow> <mo>(</mo> <mi>&omega;</mi> <mo>)</mo> </mrow> <msup> <mi>c</mi> <mi>T</mi> </msup> <mrow> <mo>(</mo> <mi>&omega;</mi> <mo>)</mo> </mrow> <msub> <mi>a</mi> <mrow> <mi>q</mi> <mo>-</mo> <mn>1</mn> </mrow> </msub> </mrow> </mfrac> <mi>d&omega;</mi> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>15</mn> <mo>)</mo> </mrow> </mrow></math>

Wherein, a_qFilter coefficients for the q-th iteration; s (ω) is the vector in (3), and

c(ω)＝[cos(φ_A，des(ω))cos(φ_A，des(ω)+ω)...cos(φ_A，des(ω)+Nω)]^T (16)

iterations may be initialized by using solutions found with previous methods as in Tkacenko et al, and by monitoring the variation of the coefficients between iterations | a_q-a_q-1‖²And when it is sufficiently small, in practice about 10^-4And stops to terminate the iteration. This approach was found to be most effective for designing IIR filters, which significantly reduces the ripple in the filter frequency response.

IIR magnitude and phase response

A feature filter method with an iterative error metric can reliably produce filters of any order. However, there is a significant performance jump that occurs at the filter order.

N＝(2h-1)·n，h≥1， (17)

Where n is the number of sampling periods corresponding to the relative time delay and h is an integer. This performance jump corresponds to the main peaks in the ideal impulse response, which can be approximated by generating a very large FIR filter using the above FIR method. The integer h end specifies the maximum number of inflection points that can occur in each band. In practice, it is useful to allow some additional sampling beyond the critical point to help minimize the pulsation amplitude, so in practice the following formula is used:

N＝(2h-1)·n+E，h≥1 (18)

where E is the additional sample. It was found that E ═ 5 gave good performance.

By design, the amplitude response is guaranteed to be flat, and any amplitude deviation is only due to numerical precision through a properly structured all-pass implementation. Fig. 15, 16 and 17 show phase responses with different values of h.

IIR filter implementation

There are many filter structures for implementing all-pass IIR filters. Most basic methods factor the filter into a series of second order parts (biquad). This is a good way to implement a general IIR filter if these parts are grouped properly. However, there is a dedicated structure that is all-pass in structure-if the coefficients are quantized, the filter is still guaranteed to be all-pass. This may lead to better numerical performance, especially in low precision fixed point implementations.

The all-pass filter lattice structure is preferred for the following reasons:

1. it is all-pass in structure so that when the coefficients are quantized, the result is still an all-pass filter.

2. It has good pointing performance. The lattice coefficients are guaranteed to be between 0 and 1, and the intermediate stages have good overflow properties.

3. It has a simple regular structure. Although it has 2 products instead of 1, which can be implemented with a straight-forward form of all-pass structure, it has a very regular multiply-accumulate structure that should be efficiently ported to a Digital Signal Processor (DSP).

Thus, an implementation where k is shown in FIG. 18₁-k_nFor the lattice coefficients from the filter table, x is one input sample and y is one output sample.

Based on IIR denominator coefficient a by using Levinson recursion₁-a_nTo find the lattice coefficient k₁-k_n. This signal flow leads to the following implementation:

a＝x-k[0]＊s[0]；

y＝s[0]+k[0]＊a；

for(i＝1；i<N；++i)

{

a＝a-k[i]＊s[i]；

s[i-1]＝s[i]+k[i]＊a；

}

s[N-1]＝a；

wherein a is an accumulator; s is a filter state array; k is a lattice coefficient.

IIR filter order reduction

The IIR group delay minimum p-th order algorithm has an advantage over the eigenfilter method in that it enables the design of more efficient filters. This is because it only uses poles in the region below the cut-off frequency (<6kHz) where the phase of the band is being modified. Above this frequency, the design method ignores the phase at higher frequencies. Fig. 23 shows a pole/zero plot for a filter designed using the group delay method.

However, for the eigenfilter approach to produce stable filters, the constraint φ must be adopted_H，des(pi) ═ N pi (as described previously). When a weight of 0 is assigned to all frequencies above the cutoff frequency, there is no way to guarantee the phase at pi. Even using small regions with non-zero weights around pi does not produce a stable filter. Thus, the algorithm distributes poles and zeros evenly around the unit circle. This allows the filter to be approximately linear in phase and gives a known phase response for all frequencies. Fig. 24 shows a pole/zero plot for a filter designed using the characteristic filter method.

It was found that some unnecessary poles and zeros can be removed after the characteristic filter algorithm has produced a stable filter. This can achieve significant filter order reduction (up to 75%) at the expense of some phase accuracy, the resulting filter no longer being approximately linear in phase at all frequencies. Since the human auditory system is insensitive to phase at higher frequencies, some phase distortion due to the removal of some poles/zeros can be tolerated, which will not become audible with respect to the unaltered filter. Fig. 25 shows the pole/zero plot for the same filter as fig. 24, but with approximately 73% of the poles/zeros removed. Fig. 27 shows the phase response before the reduction, and fig. 28 shows the phase response after the reduction.

The effect of deleting a pole close to the unit circle is mainly a local influence on the frequencies in its vicinity. However, there will be a small overall effect on all frequencies. Thus, as seen in fig. 28, deleting all high frequency poles can cause a significant phase deviation from the desired frequency response.

One way to correct for such phase deviations is to pre-distort the desired response, such pre-distortion being used in the design of the characteristic filter. Reasonable pre-twist can be found by finding the error between the reduced filter and the original filter and iteratively subtracting this error from the desired phase response.

From equations (10), (11) and (12), φ is given_H，L，des(ω)、φ_H，R，des(ω) and W (ω); let eigenfilter (phi)_H，des(ω), W (ω), N) As a function of designing a filter of length N by performing the above-described eigenfilter design method, set eiegenerfilter _ reduced (φ)_H，des(ω), W (ω), N, R) is a function that first performs a characteristic filter design and then reduces the order by a factor R by holding the lowest k poles when sorting the poles according to the addition angle, where k is given by the following equation:

to calculate the reduced corrected filter, the unreduced response of the left and right filters is first found:

a_full，L＝eigenfilter(φ_H，L，des(ω)，W(ω)，N) (20)

a_full，R＝eigenfilter(φ_H，R，des(ω)，W(ω)，N) (21)

calculating the relative phase between the left filter and the right filter:

a_{fullφrel，full}(ω)＝phase(a_full，R)-phase(a_full，L) (22)

next, several iterations are performed to pre-warp the desired phase response that is passed to the feature filter design routine. First, an initial value of an iteration with an originally expected phase response is sent;

φ_{H，L，des，0}(ω)＝φ_H，L，des(ω) (23)

φ_{H，R，des，0}(ω)＝φ_H，R，des(ω) (24)

for each iteration step i, a reduced filter is calculated based on the updated expected response:

a_i，L＝eigenfilter_reduced(φ_{H，L，des，i}(ω)，W(w)，N，R) (25)

a_i，R＝eigenfilter_reduced(φ_{H，R，des，i}(ω)，W(w)，N，R) (26)

and calculating the relative phase between the left filter and the right filter:

φ_rel，i(ω)＝phase(a_i，R)-phase(a_i，L) (27)

then, the error between the current reduced filter and the original unreduced filter is found:

Δ_i(ω)＝unwrap(φ_rel，i(ω)-φ_rel，full(ω)) (28)

this error is used to update the desired response. However, since the response above the cut-off is expected to differ, minimal modifications should be made to the response in this range, although it is desirable to avoid unnecessary discontinuities. One way to count this is to make the desired response transition linear from the last corrected frequency up to Nyquist.

<math> <mrow> <mi>C</mi> <mrow> <mo>(</mo> <mi>&omega;</mi> <mo>)</mo> </mrow> <mo>=</mo> <mrow> <mfenced open='{' close=''> <mtable> <mtr> <mtd> <msub> <mi>&Delta;</mi> <mi>i</mi> </msub> <mrow> <mo>(</mo> <mi>&omega;</mi> <mo>)</mo> </mrow> <mo>,</mo> </mtd> <mtd> <mn>0</mn> <mo>&le;</mo> <mi>&omega;</mi> <mo>&le;</mo> <mi>R</mi> <mo>&CenterDot;</mo> <mi>&pi;</mi> </mtd> </mtr> <mtr> <mtd> <mfrac> <mrow> <msub> <mrow> <mo>-</mo> <mi>&Delta;</mi> </mrow> <mi>i</mi> </msub> <mrow> <mo>(</mo> <mi>R</mi> <mo>&CenterDot;</mo> <mi>&pi;</mi> <mo>)</mo> </mrow> </mrow> <mrow> <mi>&pi;</mi> <mrow> <mo>(</mo> <mn>1</mn> <mo>-</mo> <mi>R</mi> <mo>)</mo> </mrow> </mrow> </mfrac> <mi>&omega;</mi> <mo>+</mo> <mfrac> <mrow> <msub> <mi>&Delta;</mi> <mi>i</mi> </msub> <mrow> <mo>(</mo> <mi>R</mi> <mo>&CenterDot;</mo> <mi>&pi;</mi> <mo>)</mo> </mrow> </mrow> <mrow> <mn>1</mn> <mo>-</mo> <mi>R</mi> </mrow> </mfrac> <mo>,</mo> </mtd> <mtd> <mi>R</mi> <mo>&CenterDot;</mo> <mi>&pi;</mi> <mo>&le;</mo> <mi>&omega;</mi> <mo>&le;</mo> <mi>&pi;</mi> </mtd> </mtr> </mtable> </mfenced> </mrow> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>29</mn> <mo>)</mo> </mrow> </mrow></math>

Finally, the expected response for the next iteration is generated.

<math> <mrow> <msub> <mi>&phi;</mi> <mrow> <mi>H</mi> <mo>,</mo> <mi>L</mi> <mo>,</mo> <mi>des</mi> <mo>,</mo> <mi>i</mi> <mo>+</mo> <mn>1</mn> </mrow> </msub> <mrow> <mo>(</mo> <mi>&omega;</mi> <mo>)</mo> </mrow> <mo>=</mo> <msub> <mi>&phi;</mi> <mrow> <mi>H</mi> <mo>,</mo> <mi>L</mi> <mo>,</mo> <mi>des</mi> <mo>,</mo> <mi>i</mi> </mrow> </msub> <mrow> <mo>(</mo> <mi>&omega;</mi> <mo>)</mo> </mrow> <mo>+</mo> <mfrac> <mrow> <mi>C</mi> <mrow> <mo>(</mo> <mi>&omega;</mi> <mo>)</mo> </mrow> </mrow> <mn>2</mn> </mfrac> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>30</mn> <mo>)</mo> </mrow> </mrow></math>

<math> <mrow> <msub> <mi>&phi;</mi> <mrow> <mi>H</mi> <mo>,</mo> <mi>R</mi> <mo>,</mo> <mi>des</mi> <mo>,</mo> <mi>i</mi> <mo>+</mo> <mn>1</mn> </mrow> </msub> <mrow> <mo>(</mo> <mi>&omega;</mi> <mo>)</mo> </mrow> <mo>=</mo> <msub> <mi>&phi;</mi> <mrow> <mi>H</mi> <mo>,</mo> <mi>R</mi> <mo>,</mo> <mi>des</mi> <mo>,</mo> <mi>i</mi> </mrow> </msub> <mrow> <mo>(</mo> <mi>&omega;</mi> <mo>)</mo> </mrow> <mo>-</mo> <mfrac> <mrow> <mi>C</mi> <mrow> <mo>(</mo> <mi>&omega;</mi> <mo>)</mo> </mrow> </mrow> <mn>2</mn> </mfrac> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>31</mn> <mo>)</mo> </mrow> </mrow></math>

To illustrate this approach, fig. 26 shows the raw phase response to the left and right filters that give the response shown in fig. 27. As shown in fig. 28, after the reduction, the response exhibits a significant phase deviation. To correct for the offset, the desired phase response is pre-distorted. Fig. 29 shows the phase response of the pre-twist after five iterations. This results in a corrected phase response in fig. 30.

In practice, the response will be greatly improved over eight iterations. Sometimes, after several iterative refinements, the result will deviate from the desired result, sometimes becoming unstable. Therefore, it is useful to track the quality metric through iterations and pick the iteration that performs best.

In a vehicle

Fig. 8(a, b), 9(a, b) and 11(a, b) show exemplary filters and phase responses with a difference in distance from each listening position to the two speakers of approximately 0.33 meters. Thus, the first band being phase adjusted starts at 250Hz and ends at 750Hz, respectively, with the band structure repeating every 1 kHz. While this example is found to work for many vehicle environments, the filters can be customized for a particular vehicle by measuring its appropriate internal dimensions.

Many vehicles include left and right speakers (or speaker channels) in a front passenger region of the vehicle and left and right speaker channels in a rear passenger region. Since the front passenger primarily receives sound from the front channel and the rear passenger receives sound from the rear channel, and since the distances from the passengers to the speakers may be different for the front passenger and the rear passenger, it is advantageous to apply the implementation of the invention twice, once for the front speakers heard by the front passenger and once for the rear speakers heard by the rear passenger, with each pair of filters designed using the delta distances associated with that row of speakers and the seating position. If there are additional rows of passengers, each with additional speakers, then the implementation of the invention may be repeated. The result is that each row of passengers sitting on the left and right sides of the vehicle perceives improved imaging. It should be noted that since the IDP is no longer zero for a position equidistant from the left and right speakers, i.e., a passenger seated in the center of each row of seats, the imaging deteriorates for a passenger seated in the center of the vehicle.

Multi-channel loudspeaker

Many vehicles also use multi-way speaker systems to reproduce a full range of audio frequencies. Typically, the woofer is placed at the door low, and the midrange/tweeter is placed at the door high or on the front dashboard. In these multi-way speaker configurations, the delta distance to the listener for the woofer is typically different than the delta distance for the mid/tweeter. In this case, if the crossover frequency is low enough to be within the frequency range of the band being phased, it is not possible to design a single pair of filters that work for both the low and mid/high frequency speakers. This problem can be ameliorated in a number of ways.

First, since the human auditory system is somewhat more phase sensitive at lower frequencies, the delta distance to the woofer can be used for filter design and the upper frequency limit of the phase adjusted band can be reduced approximately to the loudspeaker crossover frequency.

Second, implementations of the invention can be applied multiple times to produce separate pairs of filters tailored for each pair of low and mid/high frequency speaker pairs. In this way, each pair of low-frequency or mid/high-frequency loudspeaker pairs has a filter that adjusts only the band that falls in the frequency range of the loudspeaker, and each pair of filters is designed based on, in particular, the delta distance of the loudspeaker pair to the listener.

Surround sound

As described above, aspects of the present invention have been found to be advantageous for the sound quality of a two-channel stereo presentation in which there are listening positions that are symmetrically off-axis. Aspects of the invention also have benefits for the presentation of stereo material with more than two channels, e.g. multi-channel surround. Such applications of aspects of the invention are described next.

Four-track surround

Four-channel loudspeakers are very common, especially in the automotive market. Since a common surround formant includes the discrete signals of the center speaker, the center signal is typically combined equally with the left and right signals and rendered by the left and right speakers. Since the left and right speakers contain important common content in this case, the application of aspects of the present invention to the left and right speaker signals results in improved imaging of the center signal content.

Alternatively, aspects of the present invention may be applied only to center content before being combined with the left and right channel signals. In this way, imaging is improved for common content produced by the center channel signal, but the left and right signals are unchanged. This assumes that there is little or no common content between the left and right audio signals before they are combined with the center content.

Applying aspects of the invention to the front left and right speaker signals is important for delivering the content in the correct perceptual location. In addition, applying aspects of the present invention to the rear speakers is also advantageous for the listening experience. For content intended to come from behind the listener, particularly from 6.1 sources (such as Dolby Pro Logic IIx or Dolby Digital EX), aspects of the invention applied to the rear speakers help ensure that the rear virtual image is properly centered and the audible comb filtering effect is minimized. "Dolby", "Dolby Digital", "Dolby Pro Logic IIx" and "Dolby Digital EX" are trademarks of Dolby Laboratories Licensing Corporation.

In a vehicle, the direct path between the front speaker and the rear passenger is typically blocked by the front seat. To compensate for this, some of the front content may be mixed into the rear speakers. By applying aspects of the present invention to the rear speakers, imaging may be improved for the rear passengers in the same manner that it assists the passengers.

Five channel surround or three channel LCR rendering

Fig. 19 shows the listening position and speaker layout of the front seat of the vehicle when the left speaker, center speaker and right speaker are present. Note that the center speaker may not be on the same axis as the left and right speakers, but this may be adjusted by introducing a delay. With this configuration, the center signal appears to come from the center line of the vehicle (between listeners), rather than in front of each listener.

One previous solution to this problem is to mix some of the center channel signals into the left and right speakers and scale down the center speaker level. This solution helps to drag the central virtual image somewhat in front of each listener, since the left listener is close to the left speaker and the right listener is close to the right speaker. However, this approach is limited by the fact that it also produces significant comb filtering of the central content between the left and right speakers.

It was found that applying aspects of the invention to left and right speaker signals significantly improves the central virtual imaging in this speaker arrangement. This is shown in figure 24. The gain parameters a and b control the amount of synthesized center content that is mixed into the left and right speakers. These parameters can be controlled such that power is conserved. That is, a²+b²＝1。

Six-channel or seven-channel surround

Unlike a theater setting, when six or seven channels are used in the vehicle, they typically include three pairs of speakers plus a possible center front channel. In this case, for the same reasons as above, it was found to be advantageous to use the implementation of the aspects of the invention on each pair of loudspeakers. The common delta distance may be used to construct a filter or to maximize the effect, or each speaker bank pair may have a unique filter calculated using a unique delta distance to the nearest listener or nearest listeners who are not occluded by a seat.

Fig. 21a, 21b, 21c show three different examples of speaker/listener layouts in a vehicle.

The example in fig. 21a shows a four-channel speaker configuration with two listening positions. Since the delta distance at the listening position is different for the front and rear speaker pairs, the signals to each row of speakers can be processed using uniquely designed filter pairs.

The example in fig. 21b shows a more conventional four-channel speaker configuration with two rows of listeners. Since the front listener hears mainly the front speakers and the rear listener hears mainly the rear speakers, the implementation of aspects of the invention can be used in each row without interference from the other rows due to the shielding of the front seats and the directionality of the speakers. Furthermore, if each row has a different delta distance, the filter can be uniquely designed for each row.

The example in fig. 21c shows three rows of loudspeakers with two rows of listeners. As before, the occlusion provided by the front seat causes the front listener to hear primarily the front speaker. In this example, the middle speakers and rear speakers may have various implementations of the invention applied to improve the virtual image for the rear passenger. Since the middle and rear speakers have different delta distances to the rear listener, the middle and rear speakers may each have a unique filter pair.

Practice of

The invention can be implemented in hardware or software, or a combination of hardware and software (e.g., programmable logic arrays). Unless otherwise specified, any algorithm included as part of the invention is not inherently related to any particular computer or other apparatus. In particular, various general-purpose machines may be used with programs written in accordance with the teachings herein, or it may be more convenient to construct a more specialized apparatus (e.g., an integrated circuit) to perform the required method steps. Thus, the invention may be implemented in one or more computer programs executing on one or more programmable computer systems each comprising at least one processor, at least one data storage system (including volatile and non-volatile memory and/or storage elements), at least one input device or port, and at least one output device or port. Program code is applied to input data to perform the functions described herein and generate output information. Each such program may be implemented in a desired computer language (including machine, assembly, or high level procedural, logical, or object oriented programming languages) to communicate with a computer system. In any case, the language may be a compiled or interpreted language.

Each such computer program is preferably stored on or downloaded to a general or special purpose programmable computer readable storage medium or device (e.g., solid state memory or media, or magnetic or optical media) for configuring a computer and operating the computer when the storage medium or device is read by the computer system to perform the procedures described herein. The invention can also be considered to be implemented as a computer-readable storage medium, constructed with a computer program, where the storage medium so constructed causes a computer system to operate in a specific, predefined manner to perform the functions described herein.

A number of embodiments of the invention have been described. Nevertheless, it will be understood that various modifications may be made without departing from the spirit and scope of the invention. For example, some of the steps described herein may be ordered independently, and thus, may be performed in an order different than the order described.

Claims (12)

1. A method for reducing frequency-dependent phase differences occurring at certain listening positions relative to speakers that reproduce respective channels of a plurality of channels in a listening space, the phase differences occurring in a series of frequency bands in which the phase differences alternate between being predominantly in-phase and predominantly out-of-phase, the method comprising:

the phases in the multiple frequency bands in which the multiple channels are out of phase at such listening positions are adjusted.

2. The method of claim 1, wherein the listening space is an interior of a vehicle.

3. The method of claim 1 or claim 2, wherein adjusting the phase in the plurality of frequency bands comprises frequency bands satisfying the following condition: the width of the comb filter passband and valleys due to the phase difference at such listening positions will be larger than or equal to the critical bandwidth if no phase adjustment is applied.

4. A method according to any one of claims 1-3, wherein there are two channels, each channel being reproduced by one or more loudspeakers.

5. The method of claim 4, wherein the adjusting adds a 180 ° phase shift to the relative phase between the two channels.

6. The method of claim 5, wherein the phase on one channel is shifted by 90 ° and the phase in the other channel is shifted by-90 °.

7. A method according to claim 5 or claim 6 wherein the adjustment is effected by a bank of filters which provide a substantially flat magnitude response and a phase response which produces a combined phase response shift between the channels with alternating bands of 0 ° and 180 °.

8. The method of claim 7, wherein the filter comprises a Finite Impulse Response (FIR) filter.

9. The method of claim 7 when dependent on claim 5, wherein the filter comprises an Infinite Impulse Response (IIR) filter.

10. The method of claim 9, wherein the infinite impulse response filter is obtained using a eigenfilter method.

11. An apparatus configured to perform the method of any one of claims 1-10.

12. A computer program, stored on a computer-readable medium, for causing a computer to perform the method of any one of claims 1-10.

Images