WO2021193637A1

WO2021193637A1 - Fundamental frequency estimation device, active noise control device, fundamental frequency estimation method, and fundamental frequency estimation program

Info

Publication number: WO2021193637A1
Application number: PCT/JP2021/012001
Authority: WO
Inventors: 僚太大川
Original assignee: 株式会社トランストロン
Priority date: 2020-03-27
Filing date: 2021-03-23
Publication date: 2021-09-30
Also published as: JP7461192B2; JP2021157082A

Abstract

The present invention avoids delays occurring along with a fast Fourier transform and accurately estimates a fundamental frequency with little computation and few delays.　Input signals are sequentially acquired, and, each time a sample point in an input signal is acquired, a recurrence formula which includes the previous value of the sample point is used to calculate the amplitude spectrum of the input signal, and the peak frequency of the amplitude spectrum is extracted. Furthermore, each time a sample point is acquired, a bandwidth that includes the peak frequency of an amplitude spectrum is emphasized. Then a recurrence formula that includes the previous value of an autocorrelation function is used to calculate the autocorrelation function, and the peak frequency of the autocorrelation function is set as a fundamental frequency.

Description

Fundamental frequency estimator, active noise control device, fundamental frequency estimation method and fundamental frequency estimation program

The present invention relates to a fundamental frequency estimator, an active noise control device, a fundamental frequency estimation method, and a fundamental frequency estimation program.

Patent Document 1 describes a method of generating a note-based code representing music information, based on a step for estimating a fundamental frequency of an audio signal and obtaining a sequence of fundamental frequencies, and a sequence of the fundamental frequencies. It is disclosed to include steps to obtain a note-based chord.

JP-A-2002-82668

Since the method described in Patent Document 1 uses a fast Fourier transform, a delay due to data buffering occurs. Further, if the fast Fourier transform is to be executed for each sample in order to suppress the delay, the amount of calculation becomes large. Therefore, if the method described in Patent Document 1 is applied to a small device such as an embedded device, processing may be delayed or may not be realized due to lack of processing resources.

The present invention has been made in view of such circumstances, and is a fundamental frequency estimation device capable of avoiding the delay associated with the high-speed Fourier transform and accurately calculating the fundamental frequency of the input sound with a small amount of calculation and a small amount of delay. , Active noise control devices, and methods for estimating fundamental frequencies.

In order to solve the above problems, the basic frequency estimation device according to the present invention is, for example, a device that sequentially estimates the basic frequency of an input signal, and includes a sound collecting unit that sequentially acquires the input signal and a sound collecting unit of the input signal. A gradual DFT processing unit that calculates the amplitude spectrum of the input signal using a gradual formula that includes the previous value of the sample point each time a sample point is acquired, and a first unit that is the peak frequency of the amplitude spectrum. The previous time of the autocorrelation function for the first peak extraction unit that extracts the peak frequency, the filter unit that filters the band including the first peak frequency of the amplitude spectrum, and the signal processed by the filter unit. An autocorrelation function calculation unit that calculates the autocorrelation function using a gradual expression including a value, a second peak extraction unit that extracts the second peak frequency that is the peak frequency of the autocorrelation function, and the second peak. It is characterized by including an output unit that outputs a frequency as a basic frequency of the input signal.

According to the fundamental frequency estimation device according to the present invention, the first peak frequency (rough peak) is based on the amplitude spectrum of the input signal obtained by performing the progressive DFT (discrete Fourier transform) process on the collected input signal. Frequency) is extracted, processing is performed to emphasize the band in the vicinity of the peak frequency, the autocorrelation function is calculated, and the second peak frequency is extracted based on the autocorrelation function. As a result, the fundamental frequency of the input sound can be accurately calculated with a small amount of calculation and a small amount of delay without using the fast Fourier transform.

The first peak extraction is provided with a basic frequency enhancement unit that multiplies the amplitude spectrum calculated by the gradual DFT processing unit and the amplitude spectrum obtained by compressing the amplitude spectrum to 1 / m (m is a natural number) in the frequency direction. The unit may extract the peak frequency of the amplitude spectrum output from the basic frequency emphasis unit as the first peak frequency. Note that m is a natural number that depends on the wave tuning structure. This makes it possible to estimate the fundamental frequency more accurately.

The filter unit may perform a process of applying an inverse notch filter having the peak frequency of the amplitude spectrum as the center frequency. This makes it possible to emphasize a narrow band near the fundamental frequency with a simple configuration.

The second peak extraction unit may obtain a frequency having a maximum value in the frequency band including the maximum value of the autocorrelation function by parabolic fitting, and may use the frequency having the maximum value as the peak frequency of the autocorrelation function. As a result, the peak frequency existing between the discrete frequency indexes can be accurately estimated by utilizing the fact that the vicinity of the peak of the autocorrelation function fits well to the quadratic function.

Based on the storage unit that sequentially stores the array of the second peak frequencies extracted by the second peak extraction unit and the array of the second peak frequencies stored in the storage unit, with respect to the second peak frequency. A time-series filter unit that performs time-series filtering processing by a recurrence formula may be provided. As a result, the influence of outliers can be reduced by suppressing sudden changes in the estimated value. Further, since the filter processing is performed by the recurrence formula, the processing amount is small and the processing delay can be suppressed.

The sound collecting unit may include a tuning component enhancing unit that full-wave rectifies the signal received, and the recurrence DFT processing unit may calculate the amplitude spectrum of the signal output from the tuning component enhancing unit. .. As a result, the fundamental frequency component can be emphasized by utilizing the fact that the input signal contains a large number of fundamental frequency tuning components.

In order to solve the above problems, the active noise control device according to another aspect of the present invention is, for example, the basic frequency estimation device according to any one of the above and the opposite phase to the basic frequency of the input signal. It is characterized by including an active noise control unit that generates a signal and a sound emitting unit that outputs a signal having the opposite phase. As a result, noise can be suppressed in real time with a small amount of calculation.

In order to solve the above problems, the method of estimating the basic frequency according to another aspect of the present invention is, for example, a method of sequentially estimating the basic frequency of an input signal, the step of sequentially acquiring the input signal, and the above. Each time a sample point of an input signal is acquired, a step of calculating the amplitude spectrum of the input signal using a gradual equation including the previous value of the sample point, and a first peak frequency which is the peak frequency of the amplitude spectrum. And the step of applying a filter that emphasizes the band including the first peak frequency of the amplitude spectrum, and the previous value of the autocorrelation function for the signal in which the band including the first peak frequency is emphasized. The step of calculating the autocorrelation function using the gradual equation including the above, the step of extracting the second peak frequency which is the peak frequency of the autocorrelation function, and the second peak frequency as the basic frequency of the input signal. It is characterized by including a step to output. As a result, the fundamental frequency of the input sound can be accurately calculated with a small amount of calculation without using the fast Fourier transform.

In order to solve the above problems, in the basic frequency estimation program according to another aspect of the present invention, for example, a computer obtains a sound collecting unit that sequentially acquires the input signal and a sample point of the input signal. For each, a gradual DFT processing unit that calculates the amplitude spectrum of the input signal using a gradual formula that includes the previous value of the sample point, and a first peak frequency that is the peak frequency of the amplitude spectrum are extracted. A gradual expression that includes the previous value of the autocorrelation function for the signal processed by the 1-peak extraction unit, the filter unit that applies a filter that emphasizes the band including the first peak frequency of the amplitude spectrum, and the filter unit. The autocorrelation function calculation unit that calculates the autocorrelation function using the above, the second peak extraction unit that extracts the second peak frequency that is the peak frequency of the autocorrelation function, and the second peak frequency of the input signal. It is characterized by functioning as an output unit that outputs as a basic frequency.

According to the present invention, the delay associated with the fast Fourier transform can be avoided, and the fundamental frequency of the input sound can be accurately calculated with a small amount of calculation and a small amount of delay.

It is a figure which shows the outline of the electrical functional block of the fundamental frequency estimation device 1 and the active noise control device 2 which has this. The graph shows the state of processing in the recurrence formula DFT processing unit 12, (a) is a graph showing an example of the input audio signal, and (b) is the discrete Fourier transform of the audio signal by the recurrence formula. It is a graph which shows the amplitude spectrum obtained by. It is a graph which shows the state of the processing in the fundamental frequency emphasis part 13, (a) is a graph which shows an example of the amplitude spectrum of the signal before processing by the fundamental frequency emphasis part 13, and (b) is the graph by the fundamental frequency emphasis part 13. It is a graph which shows the amplitude spectrum of the signal after processing. It is a figure which shows typically the state of the parabolic fitting process in the 1st peak extraction part 14. It is a graph which shows the state of the reverse notch filter processing in the filter unit 15, (a) is a graph which shows an example of the amplitude spectrum of the signal before the reverse notch filter processing, (b) is the reverse notch filter which emphasizes 300kHz. It is a graph which shows the amplitude spectrum of the signal after application. It is the graph which showed the state before and after the reverse notch filter processing superposed. The graph shows the state of calculation by the autocorrelation function calculation unit 16, (a) is a graph showing an example of the autocorrelation function of the signal subjected to the inverse notch filter processing by the filter unit 15, and (b) is the inverse. It is a graph which shows an example of the autocorrelation function of the signal which has not been notch-filtered. It is a graph which shows the state of the processing in the time series filter unit 18. It is a figure which shows the outline of the electric functional block when the fundamental frequency estimation apparatus 1 and SAN are combined in ANC processing unit 21. FIG. 5 is a flowchart showing a flow in which a fundamental frequency of a signal is estimated by a fundamental frequency estimation device 1 and the active noise control device 2 reduces noise based on the fundamental frequency. It is a figure which shows the outline of the electric functional block of a general SAN.

Hereinafter, embodiments of the fundamental frequency estimation device and the active noise control device according to the present invention will be described in detail with reference to the drawings. The fundamental frequency estimation device is a device that estimates the fundamental frequency in real time by sequentially acquiring external sounds and analyzing them as input signals. The active noise control device is a device that reduces noise by outputting a signal having the opposite phase of an external sound based on the fundamental frequency estimated by the fundamental frequency estimation device.

The fundamental frequency estimation device and the active noise control device can be applied to reduce so-called engine muffled noise, for example, in an automobile, but the application is not limited to this. The fundamental frequency estimation device can be applied to, for example, a device that automatically transcribes in real time, a device that estimates pitch in real time, and a communication device for emphasizing voice tuning components and realizing hands-free communication.

FIG. 1 is a diagram showing an outline of an electrical functional block of a fundamental frequency estimation device 1 and an active noise control device 2 having the fundamental frequency estimation device 1. The fundamental frequency estimation device 1 is used as a software resource at least by a computing device such as a CPU (Central Processing Unit) for executing information processing and a storage device such as a RAM (Random Access Memory) or a ROM (Read Only Memory). Self-correlation between the sound collecting unit 10, the tuning component emphasizing unit 11, the gradual DFT (Discrete Fourier Transform) processing unit 12, the fundamental frequency emphasizing unit 13, the first peak extraction unit 14, and the filter unit 15. It has a function calculation unit 16, a second peak extraction unit 17, a time series filter unit 18, an output unit 19, and a storage unit 20.

The sound collecting unit 10 is a functional unit that sequentially acquires, for example, an external sound picked up by a microphone as an input signal.

The tuning component emphasizing unit 11 is a functional unit that full-wave rectifies the signal acquired by the sound collecting unit 10 and emphasizes the tuning component of the input signal. As a result, the fundamental frequency component can be emphasized by utilizing the fact that the input signal contains a large number of fundamental frequency tuning components. The wave tuning component emphasizing unit 11 is not an essential configuration.

The recurrence formula DFT processing unit 12 is a functional unit that calculates the amplitude spectrum of the input signal by a discrete Fourier transform (DFT, Discrete Fourier Transform). The recurrence formula DFT processing unit 12 calculates the amplitude spectrum of the input signal using the recurrence formula including the previous value of the sample point each time the sample point of the input signal is acquired. Here, the signal processed by the recurrence type DFT processing unit 12 is a signal that is full-wave rectified by the tuning component enhancing unit 11 when the tuning component enhancing unit 11 is provided, and the tuning component enhancing unit 11 is used. If not, it is a signal acquired by the sound collecting unit 10.

Hereinafter, the processing performed by the recurrence formula DFT processing unit 12 will be specifically described. The recurrence formula DFT processing unit 12 uses the following recurrence formula (1) to perform a discrete Fourier transform of the past N samples (DFT length is N) at time t each time a new sample point xt is acquired. calculate. The subscript integer k is the index of the frequency bin.

... (1)

The discrete Fourier transform of the past N samples is expressed by the following equation (2). Here, x _t , x _t-1 , ... X _{t−N + 1} are a series of past N samples at time t.

... (2)

The absolute value of the discrete Fourier transform is the amplitude spectrum, which represents the magnitude of each frequency component of the input signal. That is, each value of the amplitude spectrum is obtained by the following equation (3).

... (3)

FIG. 2 is a graph showing a state of processing in the recurrence formula DFT processing unit 12, (a) is a graph showing an example of an input audio signal, and (b) is a recurrence formula of the audio signal. It is a graph which shows the amplitude spectrum obtained by discrete Fourier transform by. As shown in FIG. 2B, by performing the recurrence DFT process, it becomes clear that the input audio signal has peaks in the vicinity of 300 Hz and in the 2nd and 3rd overtones thereof.

In general, the discrete Fourier transform when the DFT length is N is defined by the following equation (4), but if calculated according to this definition, it will be discrete after waiting for N sample points to be acquired. It is necessary to perform Fourier transform, which causes a delay corresponding to the DFT length. Further, the larger the frequency resolution, the larger the delay, and it is difficult to estimate the fundamental frequency in real time. Further, since this calculation method requires a large amount of calculation, processing time is long and the delay is further increased. Therefore, when it is applied to an active noise control device, a time lag occurs between the generated sound and the external sound, and it is difficult to effectively reduce the noise.

... (4)

On the other hand, as in the present embodiment, by performing the discrete Fourier transform based on the previous value by the recurrence formula, the delay due to waiting for the acquisition of N sample points is not generated and the processing amount is reduced. Can be done and delays can be minimized.

Return to the explanation in Fig. 1. The fundamental frequency enhancement unit 13 is a functional unit that emphasizes the fundamental frequency component of the signal. When it is known that harmonics are present, the fundamental frequency enhancement unit 13 uses the HPS method (Harmonic Product Spectrum method), for example, to obtain the fundamental frequency band with respect to the amplitude spectrum calculated by the progressive DFT processing unit 12. Adds an amount based on the harmonic content of the band to.

For example, when it is known that the fundamental frequency and its second harmonics exist, the fundamental frequency enhancement unit 13 compresses the original amplitude spectrum and the original amplitude spectrum by 1/2 in the frequency direction, as shown in the equation (5). The fundamental frequency component is emphasized by multiplying it with the obtained amplitude spectrum.

... (5)

FIG. 3 is a graph showing the state of processing in the fundamental frequency enhancement unit 13, (a) is a graph showing an example of the amplitude spectrum of the signal before processing by the fundamental frequency enhancement unit 13, and (b) is a graph showing an example of the amplitude spectrum of the signal before processing by the fundamental frequency enhancement unit 13. It is a graph which shows the amplitude spectrum of the signal after processing by the emphasis section 13. In FIG. 3A, peaks can be seen near 300 Hz, around 600 Hz, and around 900 Hz, respectively. On the other hand, in FIG. 3 (b), the peak near 300 Hz is emphasized more than the graph shown in FIG. 3 (a), and the peaks near 600 Hz and 900 Hz are smaller. In this way, by performing the processing by the fundamental frequency emphasizing unit 13, the fundamental frequency can be emphasized by using the harmonic component of the fundamental frequency. Therefore, in the processing of the fundamental frequency emphasizing unit 13 described in detail later, more accurately. The fundamental frequency can be estimated.

In the equation (4), the fundamental frequency enhancement unit 13 multiplied the original amplitude spectrum and the amplitude spectrum of the second harmonic overtone, but instead of or in addition to this, the original amplitude spectrum is multiplied by the harmonic overtone of the third harmonic overtone or more. You may perform the process of multiplying the amplitude spectrum of the component. That is, when m harmonics (m is a natural number) are present, the fundamental frequency enhancement unit 13 has an amplitude spectrum calculated by the recurrence formula DFT processing unit 12 and an amplitude obtained by compressing the amplitude spectrum to 1 / m in the frequency direction. Multiply the spectra.

Further, the fundamental frequency emphasizing unit 13 may determine whether or not a harmonic component of the fundamental frequency is present, and perform a process of emphasizing the fundamental frequency component only when it is determined that the harmonic component is present. .. Thereby, the process of emphasizing the fundamental frequency component can be appropriately performed.

Returning to the description of FIG. The first peak extraction unit 14 is a functional unit that estimates a rough fundamental frequency. The first peak extraction unit 14 estimates the peak frequency from the amplitude spectrum output from the fundamental frequency enhancement unit 13 using the following equation (6). Note that k _max is a frequency index that takes the maximum value of the amplitude spectrum.

... (6)

Note that the first peak extraction unit 14 may calculate the approximate value of the fundamental frequency more accurately by performing a parabolic fitting process instead of the process of the equation (6). The parabolic fitting process is a process of approximating the periphery of a peak with a quadratic function and calculating a more accurate peak position.

FIG. 4 is a diagram schematically showing a state of the parabolic fitting process in the first peak extraction unit 14. In the parabolic fitting process, as shown in the following equation (7), the amplitude spectrum b of the frequency index that takes the maximum value of the amplitude spectrum and the amplitude spectra a and c before and after the amplitude spectrum are extracted, and a quadratic curve passing through these is extracted. The frequency that complements and takes the maximum value is defined as the peak frequency of the amplitude spectrum.

... (7)

By using the parabolic fitting process, the peak frequencies existing between the discrete frequency indexes can be obtained more accurately.

The peak frequency extracted by the first peak extraction unit 14 (corresponding to the first peak frequency of the present invention) is a rough fundamental frequency of the input signal and is an approximate value of the fundamental frequency. 1 is different from the value output as the estimated value. The fundamental frequency estimation device 1 estimates the fundamental frequency of the input signal more accurately by a process described later.

Further, the first peak extraction unit 14 may perform a process of convolving the discrete Fourier transform result of the window function with respect to the amplitude spectrum before the peak extraction process such as parabolic fitting. As a result, the amplitude spectrum blurred by the rectangular window can be sharpened to some extent and corrected. However, this window function processing is not essential.

Return to the explanation in Fig. 1. The filter unit 15 is a functional unit that emphasizes a specific narrow band. Specifically, the filter unit 15 performs a process of applying a filter that emphasizes the band including the peak frequency of the amplitude spectrum extracted by the first peak extraction unit 14.

In the present embodiment, the filter unit 15 calculates a process of applying an inverse notch filter having the peak frequency of the amplitude spectrum as the center frequency to the input signal to the filter unit 15 by using a gradual equation. Assuming that the bandwidth constant is r (0 <r <1), the recurrence formula of the inverse notch filter is defined below.

... (8)

... (9)

Here, x _t is the input signal, y _t is the output of the reverse notch filter, and fs is the sampling frequency. The constant r is a constant indicating that the closer to 1 the emphasis is on the frequency in a narrow range, and here, it is about 0.95 to 0.98. By this processing, it is possible to emphasize a narrow band with a simple configuration.

FIG. 5 is a graph showing the state of the reverse notch filter processing in the filter unit 15, FIG. 5A is a graph showing an example of the amplitude spectrum of the signal before the reverse notch filter processing, and FIG. 5B emphasizes 300 kHz. It is a graph which shows the amplitude spectrum of a signal after applying an inverse notch filter. In FIG. 5, by the reverse notch filter processing, the peak is emphasized near the fundamental frequency of 300 Hz, and the peaks of other frequencies are suppressed.

In this way, the autocorrelation function is clean by roughly estimating the band including the fundamental frequency by the recurrence formula DFT processing unit 12 and the first peak extraction unit 14 and then emphasizing the band by the filter unit 15. The cos wave shape is obtained, and the accuracy of peak extraction is improved (detailed later).

The process performed by the filter unit 15 may be a bandpass filter process. However, the reverse notch filter processing can emphasize a narrow band more easily than the bandpass filter processing.

Return to the explanation in Fig. 1. The autocorrelation function calculation unit 16 is a functional unit that calculates an autocorrelation function for the signal processed by the filter unit 15.

The autocorrelation function ACFt (i) of length M is defined by the following equation (10).

... (10)

FIG. 6 is a calculation example of the autocorrelation function. The autocorrelation function ACFt (i) is the inner product when the signals are superposed with the i-samples shifted, and is an index showing the similarity between the two superposed signals. Therefore, the peak interval of the autocorrelation function ACFt coincides with the signal period.

In the present embodiment, the autocorrelation function calculation unit 16 calculates the autocorrelation function ACFt (i) by using the recurrence formula (11) including the previous value ACFt-1 (i) of the autocorrelation function. This makes it possible to efficiently calculate the autocorrelation function.

... (11)

FIG. 7 is a graph showing the state of calculation by the autocorrelation function calculation unit 16, and FIG. 7A is a graph showing an example of the autocorrelation function of the signal subjected to the inverse notch filter processing by the filter unit 15. b) is a graph showing an example of the autocorrelation function of the signal not subjected to the inverse notch filter processing.

Since the signal outside the band of the target frequency is suppressed by the inverse notch filter processing, the autocorrelation function of the signal subjected to the inverse notch filter processing is smoother. By smoothing the autocorrelation function, it is less likely to detect an erroneous local maximum. In addition, the fit to the quadratic function is improved, and the accuracy of parabolic fitting in the second peak extraction unit 17, which will be described later, is improved. Further, since the DC component can be removed by the inverse notch filter processing, the distortion of the autocorrelation function can be suppressed.

Return to the explanation in Fig. 1. The second peak extraction unit 17 is a functional unit that extracts the peak frequency of the autocorrelation function. The second peak extraction unit 17 obtains the frequency having the maximum value by parabolic fitting in the frequency band including the maximum value of the autocorrelation function, and determines the frequency having the maximum value as the peak frequency of the autocorrelation function (second of the present invention). Extract as (corresponding to the peak frequency).

The rough basic period obtained from the amplitude spectrum by the first peak extraction unit 14 is as shown in Eq. (12).

... (12)

The maximum position of the autocorrelation function in the vicinity of the fundamental period is obtained by parabolic fitting, and this is defined as the period T of the fundamental frequency output by the fundamental frequency estimation device 1 as an estimated value.

... (13)

... (14)

Then, an estimated value of the _{peak frequency f 0} is derived from the period T.

... (15)

In this way, the autocorrelation function is similar to the cos wave, and the vicinity of the peak of the autocorrelation function fits well to the quadratic function, so the peak position can be estimated more accurately by using parabolic fitting.

_{The peak frequency f 0} of the autocorrelation function obtained by the second peak extraction unit 17 is extracted every time a sample point is acquired, and is stored in the sequential storage unit 20 as an array of peak frequencies.

In the discrete Fourier transform process, the fundamental frequency can be obtained stably, but only a rough value can be obtained. On the other hand, the autocorrelation function processing can accurately obtain the fundamental frequency, but if an erroneous peak due to noise or overtone components is selected, there is a possibility that a numerical value significantly different from the actual fundamental frequency may be calculated. In the present embodiment, the recurrence formula DFT processing unit 12 processes the signal using the fundamental frequency band roughly obtained by the discrete Fourier transform processing, and then the second peak extraction unit 17 obtains the autocorrelation function. Therefore, the fundamental frequency can be obtained accurately and stably.

The time-series filter unit 18 is a functional unit that corrects the peak frequency of the autocorrelation function based on the difference between the peak frequency of the autocorrelation function and the previous value. The time-series filter unit 18 performs time-series filtering processing on the peak frequency of the autocorrelation function by the following recurrence formula (16) based on the arrangement of peak frequencies.

... (16)

According to equation (16), a small amount of change is reflected as it is in the current value, and an extremely large change is reflected by setting an upper limit. FIG. 8 is a graph showing a state of processing in the time series filter unit 18. In FIG. 8, the signals before the time-series filtering process (see the dotted line) and after the time-series filtering process (see the solid line) are displayed in an overlapping manner. In this way, even if the fundamental frequency is estimated incorrectly and an outlier appears in the estimation, it is possible to suppress a sudden change in the estimated value and reduce the influence of the outlier. Further, since the time-series filtering process is performed by the recurrence formula, the amount of processing is small and the processing delay can be suppressed.

The time series filter unit 18 is not essential. Further, the time-series filter unit 18 replaces the configuration for calculating the tanh function (double curve tangent function) of the difference from the previous value as described above, instead of the sign function (sign function), arctan function (inverse tangent function) or erf. Processing may be executed using a function (error function).

Further, in the present embodiment, the time series filter unit 18 performs filtering processing by a recurrence formula that saturates the amount of change, but instead of this, a least squares method, robust regression, Kalman filter, median filter, or the like is used. It can also be used.

Return to the explanation in Fig. 1. The output unit 19 outputs the peak frequency of the autocorrelation function as the fundamental frequency. The output unit 19 transmits a fundamental frequency to the ANC processing unit 21, which will be described later. Further, the output unit 19 may display the value of the fundamental frequency on the screen of the connected user interface device (not shown).

Next, the functional unit of the active noise control device 2 will be described. As shown in FIG. 1, the active noise control device 2 mainly includes an active noise control (ANC) processing unit 21 and a sound emitting unit 22 in addition to the fundamental frequency estimation device 1. In the present embodiment, the fundamental frequency estimation device 1 constitutes a part of the active noise control device 2, but the fundamental frequency estimation device 1 and the active noise control device 2 have an ANC processing unit 21 and a sound emitting unit. 22 may be configured by different devices and may be connected to each other by wire or wirelessly.

The ANC processing unit 21 is a functional unit that generates a noise canceling sound to be emitted based on the fundamental frequency of the input signal estimated by the fundamental frequency estimation device 1. The ANC processing unit 21 generates a signal having a phase opposite to that of the input signal.

For example, in an in-vehicle device, low-pitched narrow-band noise (so-called muffled sound) is generated due to engine rotation. The muffled noise is a noise close to a sine wave, and its frequency appears in the second-order component of engine rotation in the case of a 4-cylinder engine and in the third-order component of engine rotation in the case of a 6-cylinder engine. For example, when the rotation speed of a 4-cylinder engine is 3000 rpm, the fundamental frequency of the muffled sound is 100 Hz (secondary component of engine rotation). In addition, the overtone component is also generated.

Generally, the muffled sound is reduced by using SAN (Single Adaptive Notch filter) or the like. FIG. 11 is a diagram illustrating an outline of a general SAN electrical functional block. SAN acquired engine speed information from CAN (Control Area Network) 100, generated cos waves and sine waves corresponding to the engine speed, and performed adaptive processing so that the observation signal of the error microphone 101 became small. The weights ω0 and ω1 are applied and output from the speaker 102. In order to execute such feedforward type ANC processing, it is necessary to acquire a reference signal such as the engine speed. However, there are cases where the reference signal cannot be acquired due to restrictions such as cost.

In the present embodiment, by combining the fundamental frequency estimation device 1 and the SAN, the muffled sound can be reduced without a reference signal such as a CAN. FIG. 9 is a diagram showing an outline of an electrical functional block when the fundamental frequency estimation device 1 and the SAN are combined in the ANC processing unit 21. The ANC processing unit 21 acquires the fundamental frequency of noise from the fundamental frequency estimation device 1, generates cos waves and sine waves having a phase opposite to the fundamental frequency, and performs adaptive processing so that the observation signal of the error microphone 101 becomes smaller. The applied weights ω0 and ω1 are applied and output from the speaker. Therefore, it is possible to configure the active noise control device 2 that can reduce the muffled noise in the automobile even if there is no reference signal. Twice

The sound emitting unit 22 is, for example, a functional unit that transmits a signal to a speaker and converts it into sound.

FIG. 10 is a flowchart showing a flow in which the fundamental frequency of a signal is sequentially estimated by the fundamental frequency estimation device 1 and the active noise control device 2 sequentially reduces noise based on the fundamental frequency. This process is continuously performed at predetermined time intervals while the input signal is input to the fundamental frequency estimation device 1.

First, when the sound signal is picked up by the microphone and the sound collecting unit 10 (step S11), the wave tuning component is emphasized by the wave tuning component enhancing unit 11 (step S12). Next, the signal in which the tuning component is emphasized is subjected to DFT processing by a recurrence formula expression (step S13). Next, a process of emphasizing the fundamental frequency is performed by HPS processing or the like by the fundamental frequency emphasizing unit 13 (step S14). Next, the peak frequency of the signal whose fundamental frequency has been emphasized is extracted (step S15).

Next, after performing an inverse notch filter process that emphasizes the band including the fundamental frequency by the filter unit 15 (step S16), the autocorrelation function is calculated (step S17), and the peak frequency of the autocorrelation function is extracted (step S16). S18). The extracted frequencies are sequentially stored in the storage unit 20.

Next, based on the data stored in the storage unit 20, time-series filtering is performed on the signal at the peak frequency of the autocorrelation function (step S19), and outliers are suppressed. The value obtained in step S19 is the value of the fundamental frequency estimated by the fundamental frequency estimation device 1.

Next, active noise control processing is performed based on the fundamental frequency obtained by the fundamental frequency estimation device 1, that is, a signal for reducing noise is generated by emitting sound having a phase opposite to the noise from the speaker (step). S20). Next, by emitting the signal generated in step S20 from the speaker, the noise and the sound having the opposite phase are physically added, and the noise is canceled (step S21).

According to the present embodiment, since the autocorrelation function is obtained after processing the signal using the band of the fundamental frequency roughly obtained by the recurrence formula DFT processing, the fundamental frequency is accurately estimated with a small amount of calculation. can do.

In DFT processing, while the fundamental frequency can be obtained stably, only a rough value can be obtained. On the other hand, the autocorrelation function processing can accurately obtain the fundamental frequency, but since the signal of the wrong peak overtone is emphasized due to the influence of noise and harmonic components, a numerical value that is significantly different from the actual fundamental frequency is calculated. There is a risk that it will end up. On the other hand, the fundamental frequency can be obtained accurately and stably by processing the signal using the band of the fundamental frequency roughly obtained by the DFT process and then obtaining the autocorrelation function.

Further, according to the present embodiment, the DFT process and the autocorrelation function process using the recurrence formula are sequentially performed on the acquired signal to avoid the delay associated with the fast Fourier transform, and the amount of calculation and the delay. The amount can be reduced.

For example, when using the Fast Fourier Transform, there will be a delay due to data buffering. Further, although the amount of calculation is not large when the buffered data is collectively executed by the fast Fourier transform, the amount of calculation is large when the fast Fourier transform is executed for each sample. On the other hand, when the DFT process using the recurrence formula is performed, the amount of calculation can be small even if the process is executed for each sample.

Further, according to the present embodiment, the DFT and the autocorrelation function are calculated by the recurrence formula, and the method is executed every time a sample point is acquired, so that the processing delay can be suppressed.

Further, according to the present embodiment, since the fundamental frequency of noise is estimated based on the recurrence formula calculation of the DFT and the autocorrelation function, active noise control is performed even in an embedded device that can perform only processing with a small processing load. Can be done.

The fundamental frequency estimation device 1 and the active noise control device 2 according to the present embodiment have a small amount of processing and can be configured in a small size, and are therefore suitable for mounting in an embedded device such as in an automobile. It can also be used for various purposes. For example, the fundamental frequency estimation device 1 can be applied to a howling detection and howling canceling device that sequentially obtains an amplitude spectrum, detects that a specific frequency is increasing, and suppresses divergence of that frequency. Further, the fundamental frequency estimation device 1 can be applied to an automatic transcription system that estimates the fundamental frequency of a sound and automatically reproduces a score, or a tuning system of an instrument that estimates and displays the fundamental frequency of the sound of an instrument. In addition, the fundamental frequency estimation device 1 can be applied to a telephone device that selectively emphasizes only the voice tuning component by configuring a comb-shaped filter based on the estimated fundamental frequency.

Although the embodiment of the present invention has been described in detail with reference to the drawings, the specific configuration is not limited to this embodiment and includes design changes and the like within a range not deviating from the gist of the present invention. .. The functional components of the fundamental frequency estimation device 1 and the active noise control device 2 may be further classified into more components according to the processing content, or one component executes processing of a plurality of components. You may. The basic frequency estimation program may be stored in advance in a storage medium built in a device such as a computer, a storage unit in a microcomputer having a CPU, etc., and installed in the computer from there, or a semiconductor memory. , A memory card, an optical disk, a photomagnetic disk, a magnetic disk, or the like may be temporarily or permanently stored (stored) in a removable storage medium.

1: Fundamental frequency estimation device 2: Active noise control device 10: Sound collecting unit 11: Harmonizing component emphasis unit 12: Gradual DFT processing unit 13: Fundamental frequency enhancement unit 14: First peak extraction unit 15: Filter unit 16 : Autocorrelation function calculation unit 17: Second peak extraction unit 18: Time series filter unit 19: Output unit 20: Storage unit 21: ANC processing unit 22: Sound emission unit

Claims

A device that sequentially estimates the fundamental frequency of an input signal.
A sound collecting unit that sequentially acquires the input signal and
A recurrence formula DFT processing unit that calculates the amplitude spectrum of the input signal using a recurrence formula including the previous value of the sample point each time a sample point of the input signal is acquired.
A first peak extraction unit that extracts the first peak frequency, which is the peak frequency of the amplitude spectrum, and a first peak extraction unit.
A filter unit that applies a filter that emphasizes the band including the first peak frequency of the amplitude spectrum, and
An autocorrelation function calculation unit that calculates the autocorrelation function using a recurrence formula that includes the previous value of the autocorrelation function for the signal processed by the filter unit.
A second peak extraction unit that extracts the second peak frequency, which is the peak frequency of the autocorrelation function,
An output unit that outputs the second peak frequency as the fundamental frequency of the input signal, and
A fundamental frequency estimator comprising.
It is provided with a fundamental frequency enhancement unit that multiplies the amplitude spectrum calculated by the recurrence formula DFT processing unit and the amplitude spectrum obtained by compressing the amplitude spectrum to 1 / m (m is a natural number) in the frequency direction.
The fundamental frequency estimation device according to claim 1, wherein the first peak extraction unit extracts the peak frequency of the amplitude spectrum output from the fundamental frequency enhancement unit as the first peak frequency.
The fundamental frequency estimation device according to claim 1 or 2, wherein the filter unit performs a process of applying an inverse notch filter having a peak frequency of the amplitude spectrum as a center frequency.
The second peak extraction unit is characterized in that the frequency having the maximum value in the frequency band including the maximum value of the autocorrelation function is obtained by parabolic fitting, and the frequency having the maximum value is set to the peak frequency of the autocorrelation function. The basic frequency estimation device according to any one of claims 1 to 3.
A storage unit that sequentially stores the array of the second peak frequencies extracted by the second peak extraction unit, and a storage unit.
A time-series filter unit that performs a time-series filtering process by a recurrence formula on the second peak frequency based on the array of the second peak frequencies stored in the storage unit.
The fundamental frequency estimation device according to any one of claims 1 to 4, wherein the basic frequency estimation device is provided.
It is provided with a tuning component emphasizing unit that full-wave rectifies the signal received by the sound collecting unit.
The fundamental frequency estimation device according to any one of claims 1 to 5, wherein the recurrence formula DFT processing unit calculates an amplitude spectrum of a signal output from the wave tuning component emphasis unit.
The fundamental frequency estimation device according to any one of claims 1 to 6.
An active noise control unit that generates a signal with a phase opposite to the fundamental frequency of the input signal,
The sound emitting unit that outputs the signals of opposite phase and
An active noise control device characterized by being equipped with.
It is a method of sequentially estimating the fundamental frequency of the input signal.
The step of sequentially acquiring the input signal and
Each time a sample point of the input signal is acquired, a step of calculating the amplitude spectrum of the input signal using a recurrence formula including the previous value of the sample point, and
The step of extracting the first peak frequency, which is the peak frequency of the amplitude spectrum, and
A step of filtering the band including the first peak frequency of the amplitude spectrum and
For the signal in which the band including the first peak frequency is emphasized, the step of calculating the autocorrelation function using the recurrence formula including the previous value of the autocorrelation function, and
The step of extracting the second peak frequency, which is the peak frequency of the autocorrelation function, and
A step of outputting the second peak frequency as the fundamental frequency of the input signal, and
A method for estimating a fundamental frequency, which comprises.
Computer,
A sound collecting unit that sequentially acquires the input signal and
A recurrence formula DFT processing unit that calculates the amplitude spectrum of the input signal using a recurrence formula including the previous value of the sample point each time a sample point of the input signal is acquired.
A first peak extraction unit that extracts the first peak frequency, which is the peak frequency of the amplitude spectrum, and a first peak extraction unit.
A filter unit that applies a filter that emphasizes the band including the first peak frequency of the amplitude spectrum, and
An autocorrelation function calculation unit that calculates the autocorrelation function using a recurrence formula that includes the previous value of the autocorrelation function for the signal processed by the filter unit.
A second peak extraction unit that extracts the second peak frequency, which is the peak frequency of the autocorrelation function,
An output unit that outputs the second peak frequency as the fundamental frequency of the input signal, and
A fundamental frequency estimation program characterized by functioning.