JP6668306B2 - Sampling frequency estimation device - Google Patents

Description

  The present invention relates to a technique for synchronizing a plurality of signals obtained by separately sampling the same waveform.

  In recent years, recording devices that can easily perform digital recording, such as IC recorders, and devices that can perform recording simultaneously with digital recording, such as smartphones, have become widespread. Here, digital recording means recording a sound signal in the form of a sample sequence obtained by sampling a sound waveform. Using a smartphone, for example, while recording a live performance video and performance sound at a location away from the player, record the performance sound with an IC recorder placed near the player, and use the IC recorder to record the performance sound recorded by the smartphone. In some cases, the performance is replaced with a recorded performance sound (or the latter is superimposed on the former). Generally, even if the sampling frequency of all recording devices is set to be the same in the device settings, minute variations occur in the sampling frequency of each recording device. This is because the clock generator that determines the sampling frequency does not operate at the exact same clock frequency. Therefore, when digitally recording the same sound waveform separately and independently by a plurality of recording devices, even if the recording start timings are aligned, the sampling frequency varies from recording device to recording device, and the sampling timing is shifted every moment. I will. As a technique for correcting such a deviation of the sampling frequency, there is a technique disclosed in each prior art document of Patent Document 1, Non-Patent Document 1, and Non-Patent Document 2.

  Non-Patent Document 1 discloses a technique for transmitting a reference signal (pilot signal) from a transmitter and detecting and correcting a frequency shift due to a sampling frequency shift from a reference signal included in a signal received on the receiver side. I have. Patent Literature 1 discloses a correction technique in a case where a sampling frequency is different between a side that transmits a measurement signal (such as a TSP signal) and a side that receives a measurement signal when measuring transfer characteristics of a sound field. In the technique disclosed in Patent Document 1, a TSP signal is repeatedly transmitted in order to suppress the influence of noise at the time of measurement, a plurality of measured TSP signals are cut out at regular intervals, and the position of each TSP signal cut out in this way is determined. The sampling frequency shift is estimated and corrected by detecting the phase difference.

  Non-Patent Document 2 discloses a technique of correcting a deviation of a sampling frequency among a plurality of recording devices using statistical signal processing. In the technique disclosed in Non-Patent Document 2, first, a reference signal is determined for each recording signal recorded by a plurality of recording devices. Then, a signal in the case where the sampling frequency is shifted from the reference signal is statistically modeled, and a signal other than the reference signal is applied to the statistical model to estimate the sampling frequency shift.

JP 2002-101500 A

Yasumatsu Matsuoka, Yusuke Nakajima, Ken Yoshimura, "Acoustic information transmission method for acquiring information with microphone of mobile terminal", NTT Docomo Technical Journal, vol.14, No.2, 2006 Shigeki Miyabe, Nobutaka Ono, andShoji Makino, “BLIND COMPENSATION OF INTER-CHANNEL SAMPLING FREQUENCY MISMATCH WITHMAXIMUM LIKELIHOOD ESTIMATION IN STFT DOMAIN,” proc. ICASSP 2013, pp.674-678

  However, the technology disclosed in Non-Patent Document 1 and the technology disclosed in Patent Document 1 have a problem that they have many restrictions and lack versatility. For example, the technique disclosed in Non-Patent Document 1 requires a device for generating a reference signal (pilot signal), and also has a problem that a recorded signal is affected by the reference signal. On the other hand, the technique disclosed in Patent Literature 1 has a problem that the same signal cannot be used unless it is repeatedly transmitted at regular intervals. On the other hand, Non-Patent Document 2 does not have the problem of lack of versatility, but its execution requires a large amount of calculation, and the calculation time required to complete the estimation of the sampling frequency shift is long. There is a problem taken.

  The present invention has been made in view of the above-described problem, and it is possible to realize synchronization of a plurality of signals obtained by separately sampling the same waveform in a shorter calculation time than before, and An object is to provide a technology having high versatility.

  In order to solve the above-described problems, the present invention provides a method in which one of a plurality of signals obtained by separately sampling the same waveform is used as a reference signal, and one of the remaining signals is used as a correction target signal. A time shift amount calculating unit that calculates a correlation between both signals for each frame while shifting one of the reference signal and the correction target signal in the time axis direction, and calculates a time shift amount of both signals for each frame according to the calculation result. And an error calculation unit that calculates, for each frame, a first estimated value that is an estimated value of an error of the sampling frequency of the correction target signal in each frame from the time shift amount calculated by the time shift amount calculation unit. Statistical processing is performed on the first estimated value calculated for each frame by the error calculating unit to calculate a second estimated value that is an estimated value of an error of a sampling frequency over the entire correction target signal. Providing sampling frequency estimating apparatus, characterized by having a statistical processing unit for outputting. Since the error in the sampling frequency of the correction target signal is a deviation of the sampling frequency of the correction target signal from the sampling frequency of the reference signal, the sampling frequency of the correction target signal can be obtained from the error and the sampling frequency of the reference signal. it can. Therefore, calculating the estimated value of the error (that is, estimating the error) is equivalent to estimating the sampling frequency of the correction target signal.

  According to such a sampling frequency estimation device, one of a plurality of signals obtained by independently sampling the same waveform is used as a reference signal, and each of the remaining signals is used as a correction target signal, and By calculating an estimated value of the sampling frequency error and correcting the error using an existing technique such as time axis companding, it becomes possible to synchronize each correction target signal with the reference signal. The sampling frequency estimating apparatus of the present invention does not require a pilot signal and does not require each signal to be repeatedly output over a certain period of time. Versatility is higher than. Although the details will be described later, according to the sampling frequency estimating apparatus of the present invention, the error of the sampling frequency of the correction target signal can be calculated in a shorter calculation time than when the technique disclosed in Non-Patent Document 2 is used. In addition, synchronization of a plurality of signals obtained by independently sampling the same waveform can be realized in a shorter calculation time than before.

  As a specific configuration of the statistical processing unit, it is estimated from the first estimated value calculated for each frame by the error calculating unit (that is, an estimated value of a rough error in each frame) that the error calculating unit statistically includes many errors. First statistical filter processing for removing outliers to be performed, and filter processing for smoothing a group of first estimated values from which outliers have been removed by the first statistical filter processing (for example, processing for calculating an average value) ) And a second statistical filtering process comprising one of a filtering process (for example, a process of selecting a median value) for selecting a representative value from the group of first estimated values, and A configuration in which the processing result of the second statistical filter processing is output as a second estimated value (estimated value of the error of the sampling frequency over the entire signal to be corrected) may be considered.

  As a specific example of the first statistical filter processing, when the first estimated values calculated for each frame by the error calculating unit are sorted in the order of their sizes, a predetermined number from each end, or a predetermined number from each end. There is a process of removing a value that does not belong to a range determined according to the value as an outlier. For example, if the predetermined number is 1/4 of the total number of the first estimated values calculated by the error calculating unit, the value smaller than the first quartile and the value larger than the third quartile are outliers. Will be excluded. Further, in a mode in which the first quartile and the third quartile are weighted to determine the above range, outliers are excluded by the so-called quartile range method.

  In a more preferred aspect, the time shift amount calculation unit excludes a frame in which the powers of the reference signal and the correction target signal are less than a predetermined threshold from calculation targets of the time shift amount. If the threshold value is set to an appropriate value, a frame that does not include the reference signal with sufficient intensity or a frame that does not include the correction target signal with sufficient intensity is excluded from the calculation target of the time shift amount. Even if the amount of time lag is calculated with reference to a frame that does not include the reference signal with sufficient strength or a frame that does not include the correction target signal with sufficient strength, the amount of error will include many errors. The estimated value of the error calculated based on such a time lag amount is likely to be excluded as an outlier in the first statistical filter processing, and the calculation of the time lag amount itself is wasted in the first place. According to such an aspect, it is possible to avoid a wasteful calculation being performed in the time lag amount calculating unit, further shorten the processing time required for estimating the sampling frequency of the correction target signal, and reduce the sampling frequency error with high accuracy. Can be calculated by

  In another preferred aspect, the time lag amount calculating unit sets the value representing the correlation between the reference signal and the correction target signal (for example, the maximum value among a plurality of cross-correlation functions calculated while shifting the time) to a predetermined value. It is characterized in that frames below the threshold value are excluded from the calculation of the amount of time shift. If the threshold value is set to an appropriate value, the time lag amount is not calculated for a frame that has a low correlation with the frame corresponding to the reference signal among the frames forming the correction target signal. Even if the amount of time shift is calculated for such a frame, it will contain many errors, and the estimated value of the error calculated based on the amount of time shift is regarded as an outlier as the first statistical filter processing. Is highly likely to be excluded, and the calculation itself of the time lag amount is wasted in the first place. According to such an aspect as well, it is possible to calculate the sampling frequency with high accuracy while avoiding unnecessary calculation being performed in the time lag amount calculating unit and further reducing the processing time required for estimating the sampling frequency of the correction target signal. Becomes possible.

  As another mode for solving the above-described problem, a mode is provided that provides a program that causes a general computer such as a CPU (Central Processing Unit) to function as the time shift amount calculation unit, the error calculation unit, and the statistical processing unit. Conceivable. By operating a general computer according to such a program, the computer can function as the sampling frequency estimation device of the present invention. Note that specific provision modes of such a program include a mode in which the program is written and distributed on a computer-readable recording medium such as a CD-ROM (Compact Disk-Read Only Memory), and a mode in which electric communication such as the Internet is used. A mode of distributing by downloading via a line is conceivable.

  Further, as still another mode for solving the above-mentioned problem, one of a plurality of signals obtained by separately sampling the same waveform is used as a reference signal, and one of the remaining signals is corrected. As a target signal, a cross-correlation function of both signals is calculated for each frame while shifting one of the reference signal and the correction target signal in the time axis direction, and a time shift amount of both signals is calculated for each frame according to the calculation result. Calculating a time shift amount to be calculated; and an error calculating step of calculating, for each frame, an estimated value of a sampling frequency error of the correction target signal in each frame from the time shift amount calculated in the time shift amount calculation step. And performing statistical processing on the estimated value of the error calculated for each frame in the error calculating step to calculate an estimated value of the error of the sampling frequency of the correction target signal. Aspect to provide a sampling frequency estimation method characterized by comprising: a statistical processing step of force, also conceivable. It is also conceivable to provide a program for causing a general computer such as a CPU to execute each of the time shift amount calculation step, the error calculation step, and the statistical processing step.

1 is a block diagram illustrating a configuration example of a sampling frequency estimation device 10 according to an embodiment of the present invention and a configuration example of a signal processing system 1 including the same. FIG. 3 is a diagram showing a result of estimating a sampling frequency for each frame obtained by a sampling frequency estimating apparatus 10. It is a figure for demonstrating a quartile and a quartile range. It is a figure for explaining an effect of outlier removal processing based on a quartile. FIG. 11 is a diagram showing experimental results of evaluation experiments on the present embodiment and the technology disclosed in Non-Patent Document 2. FIG. 11 is a diagram showing experimental results of evaluation experiments on the present embodiment and the technology disclosed in Non-Patent Document 2.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.
(A: Configuration)
FIG. 1 is a block diagram illustrating a configuration example of a sampling frequency estimation device 10 according to an embodiment of the present invention and a configuration example of a signal processing system 1 including the sampling frequency estimation device 10. The signal processing system 1 includes sound signals (samples) obtained by separately and independently sampling the same sound waveform by N (N is a natural number of 2 or more) recording devices (for example, a smartphone or an IC recorder). Column) X _n (t) (n = 1 to N) is input. Note that the existing technology may be appropriately used for synchronizing the recording start timings in the N recording devices. For example, if each recording device can execute communication via a telecommunication line such as the Internet, the recording start timing may be adjusted by the communication, and communication by other communication means such as Bluetooth (registered trademark) is possible. For example, the recording start timing may be adjusted by communication by the communication means.

  Although the sampling frequency of each of the N recording devices is set to the same value (for example, 44.1 kHz), the clock generator of each recording device does not operate at completely the same clock frequency. There is a slight deviation in the sampling frequency of the recording device. For this reason, even if each recording device performs recording at the same recording start timing and reproduces the N sound signals with their heads aligned, the deviation gradually increases and the deviation increases as the reproduction progresses. The signal processing system 1 according to the present embodiment is for estimating and correcting the sampling frequency error between the N sound signals so that the N sound signals can be synchronized.

  As shown in FIG. 1, the signal processing system 1 includes a sampling frequency estimating device 10 and a time axis companding device 20. The sampling frequency estimating device 10 is supplied with the N sound signals. The sampling frequency estimating apparatus 10 sets one of the N sound signals as a reference signal, and sets each of the remaining N-1 sound signals as a correction target signal. Is estimated for each correction target signal, and data indicating the estimation result is given to the time axis companding device 20. The time axis companding device 20 performs time axis companding on each correction target signal such that an error in the sampling frequency estimated for each correction target signal is eliminated. Thereby, synchronization of N sound signals is realized. As a time axis companding algorithm in the time axis companding apparatus 20, an existing technique may be appropriately used. In the present embodiment, the sampling frequency estimating apparatus 10 performs a process that prominently illustrates the features of the present embodiment, thereby realizing estimation of the sampling frequency error of each of the correction target signals in a shorter calculation time than before. , High versatility can be ensured. In the following, description will be made mainly on the sampling frequency estimating apparatus 10 which remarkably shows the features of the present embodiment.

  As shown in FIG. 1, the sampling frequency estimating apparatus 10 includes a short-time Fourier transform (in FIG. 1, denoted as “STFT”) unit 100, a time shift amount calculating unit 110, an error calculating unit 120, and a statistical processing unit 130. Contains. Each unit illustrated in FIG. 1 may be a hardware module configured by an electronic circuit, or may be a software module realized by operating a CPU (Central Processing Unit) according to a signal processing program.

The STFT unit 100 divides each of the sound signals X _n (t) (n = 1 to N) input to the sampling frequency estimation device 10 into frames each having a predetermined number of samples, and performs a short-time Fourier transform for each frame. The signal is converted into a signal X _n (f) in the frequency domain (where f is a variable representing the frequency, the same applies hereinafter) and given to the time shift amount calculation unit 110. A well-known conversion algorithm may be appropriately used for the STFT unit 100.

The time shift amount calculation unit 110 selects one of the N sound signals as a reference signal, sequentially selects each of the remaining N-1 sound signals as a correction target signal, and selects the reference signal and the correction target. The cross-correlation function of both signals is calculated for each frame while shifting one of the signals in the time axis direction, and the time shift amount of both signals is calculated for each frame. In the following, the process executed by the time shift amount calculation unit 110 will be described in detail by taking as an example a case where X _ref (f) is selected as the reference signal and X _k (f) (k ≠ ref) is selected as the correction target signal. Will be described.

The time shift amount calculation unit 110 first changes the cross-correlation function C (τ) of the reference signal X _ref (f) and the correction target signal X _k (f) (k） ref) for each frame by changing the value of τ. While calculating. The reason for calculating the cross-correlation function C (τ) is that a method for estimating a sample point at which the cross-correlation function is maximized is generally known as a method for correcting the time shift amount. Generally, when there are two analog signals x (t) and y (t) in the time domain, the cross-correlation function C (τ) is expressed by the following equation (1). In the case of a digital signal, it is expressed by the following equation (2).

The cross-correlation function C (τ) calculated by Equation 1 or 2 corresponds to taking an inner product by shifting one of the two signals by τ in the time axis direction. If the assumption that “the same signal is included somewhere in the two signals” holds, it is considered that the time shift amount of the two signals can be estimated by obtaining τ at which the cross-correlation function is maximum. . Considering the Fourier transforms X (f) and Y (f) of the two signals, the cross-correlation function C (τ) is calculated by the following equation (3). IFFT () on the right side of Expression 3 is an operator representing an inverse Fourier transform, and X ^* (f) represents a complex conjugate of X (f).

The time shift amount calculation unit 110 according to the present embodiment changes the cross-correlation function C (τ) of each frame of the reference signal X _ref (f) and the correction target signal X _k (f) (k ≠ ref) by changing the value of τ. While calculating according to Equation 3. Specifically, the time shift amount calculation unit 110 sets the complex conjugate of the signal X _ref (f) for the i-th frame in the reference signal to X ^* (f) on the right-hand side of Expression 3, and sets the i-th frame in the correction target signal. Using the signal X _k (f) (k （ref) for the frame as Y (f) on the right-hand side of Formula 3, the calculation of Formula 3 is performed while changing τ, and τ at which the cross-correlation function C (τ) is maximized is specified. . Then, the time shift amount calculation unit 110 converts the τ specified in this manner into the estimated value N _ki of the time shift amount for the i-th frame of the correction target signal X _k (f) (that is, the cross-correlation function C ( is given to the error calculation unit 120 as τ) that maximizes τ). The same applies to frames of other numbers.

  In this embodiment, the calculation of the cross-correlation function C (τ) by the calculation in the frequency domain shown in Expression 3 instead of the calculation in the time domain of Expression 1 or 2 is advantageous in terms of calculation amount. Because there is. In the present embodiment, the STFT unit 100 is provided so that the time shift amount calculation unit 110 can calculate the cross-correlation function C (τ) by the calculation shown in Expression 3. Therefore, the STFT unit 100 may be omitted if the time shift amount calculation unit 110 calculates the cross-correlation function C (τ) by the calculation shown in Expression 1 or Expression 2.

The error calculating unit 120 estimates the error of the sampling frequency error of the correction target signal with respect to the sampling frequency f _s of the reference signal based on the time shift amount N _ki of each frame given from the time shift amount calculating unit 110 (hereinafter, referred to as the (Estimated value of 1) E _ki is calculated for each frame. For example, if the time shift amount of the i-th frame is N _ki and the first sample of the i-th frame in the correction target signal X _k (f) is the S _ki- th sample from the head of the signal, , The error calculation unit 120 calculates the first estimated value E _ki for the i-th frame according to the following _Expression 4, and gives the calculated value to the statistical processing unit 130. As described above, since the time deviation amount N _ki calculated by the time shift amount calculation unit 110 is a shift amount relative to the beginning of the reference signal, a cross correlation function for each frame by STFT as in this embodiment If found, it is the amount of displacement based on the top of the frame. For this reason, the estimated value E _ki of the error of the sampling frequency of the correction target signal X _k (f) based on the cross-correlation function in the i-th frame is represented by the following _Expression 4.

The statistical processing unit 130 performs statistical processing on the first estimated value E _ki calculated for each frame by the error calculating unit 120, and estimates the sampling frequency error over the entire signal to be corrected (hereinafter, referred to as a second estimated value). Value) E is calculated and output to the time axis compander 20. As shown in FIG. 1, the statistical processing unit 130 includes a first statistical filter processing unit 130a and a second statistical filter processing unit 130b. In other words, the statistical processing performed by the statistical processing unit 130 includes the processing performed by the first statistical filter processing unit 130a and the processing performed by the second statistical filter processing unit 130b. The contents of the processing executed by each of these statistical filter processing units are as follows.

The first statistical filter processing unit 130a _removes, from the first estimated value E _ki calculated for each frame by the error calculating unit 120, an outlier that is statistically estimated to include a large number of errors. Perform filter processing. The first estimated value E _ki calculated for each frame by the error calculator 120 often includes many errors. The first estimated value E _ki is based on only information on the i-th frame of each of the correction target signal X _k (f) and the reference signal X _ref (f), and roughly determines the difference between the sampling frequencies of both signals. This is because it is an estimated value. FIG. 2 shows a sampling frequency estimation result for each frame when an experiment was performed by artificially shifting the sampling frequency by 3 Hz. If the error of the sampling frequency of the correction target signal with respect to the sampling frequency of the reference signal is known, the sampling frequency of the correction target signal can be calculated from the sampling frequency of the reference signal and the error. The estimation of the sampling frequency is equivalent. As shown in FIG. 2, it is considered that the reason why there is a large variation in the sampling frequency estimated for each frame is that it is difficult to estimate a very small time shift caused by the sampling frequency shift with high accuracy.

In the present embodiment, a process of removing, as an outlier, a value that greatly deviates from other first estimated values E _ki calculated by the error calculating unit 120 as an outlier is used as a first statistical filter process. Has been adopted. Specifically, in the present embodiment, a process based on a so-called quartile is employed as the first statistical filter process. Here, the quartile refers to the number of divisions into which the data to be processed is sorted into quartiles after they are sorted in the order of size, and the first quartile, the second quartile, and the The quantile is called the third quartile (see FIG. 3). The difference between the first quartile and the second quartile is called an interquartile range (IQR). The quartile range is one indicator of the degree of sample variation.

More specifically, the first statistical filter processing unit 130a first sorts the first estimated values E _ki calculated for each frame by the error calculating unit 120 in the order of their sizes. Next, the first statistical filter processing unit 130a determines, among the first estimated values E _ki calculated for each frame by the error calculating unit 120, a value smaller than the first quartile in the sorting result, or A value larger than the quartile is excluded as an outlier, and the rest (that is, a group of first estimated values E ′ _ki not including the outlier) is transferred to the second statistical filter processing unit 130b. Here, the operation o () for detecting an outlier is represented by the following Expression 5. Specifically, each of the first estimated values E _ki calculated by the error calculating unit 120 is substituted for e (n) in Expression 5, and if the value of the operation o () is 1, the first estimated value For example, the value E _ki is excluded as an outlier. q _L and q _H represent the first and third quartiles, respectively.

In the present embodiment, a process based on a quartile is used as the first statistical filter process, but a process using a quartile range in addition to the quartile may be used. Specifically, the operation o () shown in Expression 6 may be performed instead of the operation o () shown in Expression 5 as an operation for identifying whether or not the value is an outlier. The calculation shown in Equation 6 means adding or subtracting the IQR value to the first and third quartiles with weight. Assuming α = 0, Equation 6 matches Equation 5. A method of calculating at α = 1.5 is widely known. For example, a portion corresponding to the upper and lower whiskers in the box plot shown in FIG. 4 is calculated by this.

The second statistical filter processing unit 130b selects a representative value from a group of first estimated values E ′ _ki from which outliers have been excluded by the first statistical filter processing unit 130a. the performs filtering) for selecting the median value, gives the time scale modification apparatus 20 the processing result as a second estimate E _k. Note that a maximum value, a minimum value, or the like may be used as the representative value, but a median value is considered to be most preferable. In addition, as a second statistical filtering process performed by the second statistical filtering unit 130b, a group of first estimated values E ′ _ki excluding outliers from the first statistical filtering unit 130a are smoothed. Filter processing (processing of calculating an average value of a group of first estimated values E ′ _ki excluding outliers from the first statistical filter processing unit 130 a) may be employed, but the present applicant has performed the processing. According to the experiment, better results were obtained when the filter processing for selecting the median was employed. For this reason, in the present embodiment, filter processing for selecting a median value is employed.
The above is the configuration of the sampling frequency estimation device 10.

(B: Effect of the embodiment)
According to the present embodiment, one of the N sound signals is set as a reference signal, and each of the remaining N-1 sound signals is set as a correction target signal, and the sampling frequency of the correction target signal with respect to the sampling frequency of the reference signal is set. Is estimated by the sampling frequency estimating apparatus 10 for each correction target signal, and time axis companding is performed on the correction target signal so as to eliminate the error, thereby realizing synchronization of the N sound signals. In order to evaluate the effect of the present embodiment, the present applicant compares the technique disclosed in Non-Patent Document 2 with the estimation performance of the sampling frequency error and the calculation speed (calculation of the estimated value of the sampling frequency error). An evaluation experiment was performed from the viewpoint of the length of calculation time required to complete). The outline of this evaluation experiment is as follows.

  First, a sound signal of 10 commercially available 16-bit songs (genre is pops, each song has a time length of 10 seconds) sampled at a sampling frequency of 44.1 kHz is used as an original signal, and the original signal is used as a reference signal. A signal obtained by artificially performing resampling (± 5 Hz) on the original signal is used as a correction target signal, and an error in the sampling frequency of each correction target signal is disclosed in the sampling frequency estimating apparatus 10 of this embodiment and Non-Patent Document 2. Estimated by technology. In this evaluation experiment, a computer having Corei7 3770 driven by 3.4 GHz as a CPU and a computer having a RAM of 32 GB was used as the sampling frequency estimating apparatus 10, and MATLAB (registered trademark) was used for mounting each unit such as the STFT unit 100. Was used. MATLAB (registered trademark) is numerical analysis software from The MathWorks, USA. Similarly, the method disclosed in Non-Patent Document 2 was implemented in C / C ++ and MATLAB (registered trademark) on the same computer and executed. The FFT length is 4096 samples, a Hamming window having a window length of 4096 samples is used as an analysis window, and a shift size (ie, an update unit of τ) for calculating a cross-correlation function C (τ) is 8192. , 4096, 2048, and 1024 samples, and the data range to be used was (3/8) × T to (5/8) × T (T is the number of data).

  FIG. 5A is a diagram illustrating an experimental result regarding the estimation performance for the present embodiment, and FIG. 5B is a diagram illustrating an experimental result regarding the estimation performance for the method disclosed in Non-Patent Document 2. As is clear from a comparison between FIG. 5A and FIG. 5B, the technology disclosed in Non-Patent Document 2 exceeds the technology at the highest performance (that is, the estimation error is small). However, for example, a practical range in which recording is performed for 2 hours (7200 seconds) and the time lag between the corrected signal to be corrected and the reference signal is suppressed to 5 ms or less (the estimation error of the sampling frequency is suppressed to within 0.03 Hz). Is also achieved in the present embodiment. Therefore, even in the present embodiment, no problem occurs in a practical range. Also, from FIG. 5A, it can be seen that in the present embodiment, the same estimation performance can be realized regardless of the shift size. The shift size affects the amount of calculation. In other words, the experimental results in FIG. 5A indicate that the present embodiment can sufficiently achieve a practical range of performance even with a small amount of calculation.

  FIG. 6A is a diagram illustrating an experimental result regarding the calculation speed for the present embodiment, and FIG. 6B is a diagram illustrating an experimental result regarding the calculation speed for the method disclosed in Non-Patent Document 2. You. As is clear from comparison between FIGS. 6A and 6B, the method of the present embodiment is overwhelmingly faster (the estimation of the sampling frequency shift is completed) as compared with the method disclosed in Non-Patent Document 2. The calculation time required to perform the method is short), and it can be seen that even with the implementation using MATLAB (registered trademark), a calculation speed that exceeds the implementation of the method disclosed in Non-Patent Document 2 using C / C ++ is obtained. Summarizing the above experimental results, it is concluded that according to the present embodiment, it is possible to achieve a practical range of estimation performance in a shorter calculation time than the technique disclosed in Non-Patent Document 2.

  As described above, according to the present embodiment, synchronization of a plurality of sound signals obtained by separately sampling the same sound waveform can be realized in a shorter calculation time than the technique disclosed in Non-Patent Document 2. It becomes possible. In addition, in the present embodiment, synchronization can be performed only with a sampled sound signal (in other words, a recorded sound signal), and a pilot signal is not required. It has high versatility. In the present embodiment, each sound signal to be synchronized does not need to be repeatedly transmitted, and has higher versatility than the technology disclosed in Patent Document 1. That is, according to the present embodiment, it is possible to achieve synchronization of a plurality of signals obtained by independently sampling the same waveform in a shorter calculation time than before, and to realize high versatility. Will be possible.

(C: deformation)
Although one embodiment of the present invention has been described above, the following modifications may be made to this embodiment.
(1) In the above embodiment, a case has been described where a plurality of signals input to the sampling frequency estimating apparatus 10 are a plurality of sound signals obtained by independently sampling the same sound waveform. However, the plurality of signals input to the sampling frequency estimating apparatus 10 need only be obtained by independently sampling the same waveform, and are not limited to sound signals. Further, in the above embodiment, the process using the quartile is adopted as the first statistical filter process. For example, after the estimated values calculated for each frame by the error calculating unit 120 are sorted in the order of their sizes, A process may be adopted in which these are divided into three equal parts, and a value smaller than the value corresponding to the smaller break position and a value larger than the value corresponding to the larger break position are set as outliers. The point is that when the first estimated values calculated for each frame by the error calculating unit 120 are sorted in the order of their sizes, the first estimated values belong to a predetermined number from both ends or a range determined according to each predetermined number-th value from both ends. Any processing may be used as long as the processing is performed for outliers to be outliers.

(2) The statistical processing executed by the statistical processing unit 130 of the above embodiment is a processing based on a deterministic approach, and includes a first statistical filter processing for removing outliers using a quartile method or the like, A representative value (median value in the above embodiment) is selected from the processing result of the first statistical filtering process, and the selected value is used as an estimated value of the error of the sampling frequency over the entire signal to be corrected. Was composed. However, even if the process of calculating the second estimated value by statistically modeling the first estimated value calculated for each frame by an exponential function family and estimating the model parameters is adopted as the above-described statistical process, good. Specifically, for example, the above-described modeling is performed using the Laplace distribution, the shape of the distribution is determined by estimating the parameters of the Laplace distribution, the mode is obtained from the determined distribution, and the mode is calculated as the second mode. By estimating the estimated value, the value of the sampling frequency error is also likely to be estimated.

(3) The time lag amount calculating unit 110 of the above embodiment calculates the time lag amount based only on τ at which the cross-correlation function C (τ) becomes the maximum, but the time lag amount is calculated in descending order of the cross-correlation function C (τ). M τ may be left as a candidate, and the time shift amount may be calculated based on the M τ. For example, the time shift amount is calculated from the average value of these M τs. In addition, all τs whose values of the cross-correlation function C (τ) are equal to or larger than a predetermined threshold may be set as candidates for the amount of time shift. In this case, in order to avoid the influence of the magnitude of the power, a normalized cross-correlation function may be used.

(4) The time lag amount calculating unit 110 may exclude frames whose powers of the reference signal and the correction target signal are less than a predetermined threshold from calculation targets of the time lag amount. If the threshold is set to an appropriate value, frames that do not include the reference signal with sufficient strength or frames that do not include the correction target signal with sufficient strength are excluded from the calculation of the time shift amount. A frame that does not include the reference signal with sufficient strength or a frame that does not include the correction target signal with sufficient strength has a small contribution to the estimation of the sampling frequency shift in the first place, and even if the time shift amount is calculated for such a frame. It will contain many errors. The first estimated value calculated based on such a time shift amount is likely to be excluded as an outlier by the first statistical filter processing unit 130a as an outlier, and the calculation itself of the time shift amount is useless in the first place. Become. According to such an embodiment, it is possible to calculate the error of the sampling frequency with high accuracy while avoiding the useless calculation in the time shift amount calculation unit 110.

(5) The time lag amount calculating unit 110 may exclude a frame in which the maximum value of the cross-correlation function C (τ) is smaller than a predetermined threshold from the calculation target of the time lag amount. If the threshold is set to an appropriate value, the calculation of the amount of time lag will not be performed based on the cross-correlation function below the threshold. Even if the time lag amount is calculated based on the cross-correlation function below the threshold value, the time lag amount will include many errors, and the first estimated value calculated based on such a time lag amount is excluded as an outlier. It is highly probable that the calculation of the time lag amount itself is useless in the first place. According to such an embodiment as well, it is possible to calculate the sampling frequency with high accuracy while avoiding useless calculation in the time lag amount calculating unit.

(6) In the above embodiment, the frame size when dividing the reference signal and the correction target signal into frames is fixed. However, in such a mode, as the frame number increases, the sample shift between the two signals increases, and the cross-correlation function It is conceivable that the calculation of C (τ) becomes meaningless (or the cross-correlation function C (τ) cannot be calculated). Therefore, the processing may be stopped when the maximum value of the cross-correlation function C (τ) falls below a predetermined threshold, and some notification may be given to the user of the sampling frequency estimating apparatus 10, or the frame size may be increased. The frame of the reference signal and the frame of the signal to be corrected may be re-divided.

(7) In the above embodiment, the amount of time lag between both signals in each frame is calculated by calculating the cross-correlation function between the reference signal and the signal to be corrected for each frame. However, it is of course also possible to calculate the cross-correlation between the two signals after normalizing the two signals and calculate the amount of time shift based on the calculation result. Further, a process of calculating a difference signal between the two signals while shifting one of the reference signal and the correction target signal in the time axis direction with respect to the other is executed for each frame, and the maximum value (or power) of the amplitude of the difference signal is determined. The signal may be calculated as a value representing the correlation between the two signals, and the time lag amount of each signal may be calculated for each frame based on the calculation result, or the ratio between the sum signal and the difference signal of the two signals may be calculated. A value representing the correlation between the signals may be calculated for each frame, and the amount of time lag between the two signals for each frame may be calculated based on the calculation result. Alternatively, a value indicating the correlation between the reference signal and the correction target signal may be calculated for each frame by pattern matching, and the amount of time lag between the two signals for each frame may be calculated based on the calculation result. The point is that the correlation between the reference signal and the signal to be corrected is calculated for each frame, and the time shift amount of each signal for each frame is calculated according to the calculation result.

DESCRIPTION OF SYMBOLS 1 ... Signal processing system, 10 ... Sampling frequency estimation apparatus, 20 ... Time axis companding apparatus, 100 ... STFT section, 110 ... Time deviation amount calculation section, 120 ... Error calculation section, 130 ... Statistical processing section, 130a ... First , A statistical filter processing unit, 130b... A second statistical filter processing unit.

Claims (4)

An error calculating step of calculating, for each frame, an error of a sampling frequency of a correction target signal being a second sound signal with respect to a sampling frequency of a reference signal being a first sound signal;
A statistical processing step of performing statistical processing on the error calculated for each frame in the error calculating step and estimating a sampling frequency error over the entire correction target signal;
A synchronization step of performing time-axis companding on the correction target signal so that the error estimated in the statistical processing step is eliminated and synchronizing the correction target signal with the reference signal,
The synchronization method, wherein the reference signal and the correction target signal are sound signals obtained by separately and independently sampling the same sound waveform. Time shift calculation for calculating the correlation of both signals for each frame while shifting one of the reference signal and the correction target signal in the time axis direction, and calculating the time shift of both signals for each frame according to the calculation result. Including steps
In the error calculation step, an error of the sampling frequency of the correction target signal in each frame is calculated for each frame from the time shift amount calculated in the time shift amount calculation step.
The method according to claim 1, wherein: Error calculating means for calculating, for each frame, an error of a sampling frequency of a correction target signal which is a second sound signal with respect to a sampling frequency of a reference signal which is a first sound signal;
Statistical processing means for performing statistical processing on the error calculated for each frame by the error calculating means and estimating a sampling frequency error over the entire correction target signal;
Synchronizing means for performing time-axis companding on the correction target signal so as to eliminate the error estimated by the statistical processing means and synchronizing the correction target signal with the reference signal,
A synchronizer, wherein the reference signal and the correction target signal are sound signals obtained by separately and independently sampling the same sound waveform . Time shift calculation for calculating the correlation of both signals for each frame while shifting one of the reference signal and the correction target signal in the time axis direction, and calculating the time shift of both signals for each frame according to the calculation result. Including means,
The error calculating means calculates, for each frame, an error of the sampling frequency of the correction target signal in each frame from the time lag amount calculated by the time lag amount calculating means.
The synchronizer according to claim 3, wherein:

Images