JPH0887298A - Method and device for processing audio signal - Google Patents

Abstract

PURPOSE: To correctly detect a pitch with an audio signal of a small bit length and to attain correct pitch correction processing by using the pitch. CONSTITUTION: An input digital audio signal of 24 bits is stored in a pitch operation data memory 26 while preserving respective codes as the quadripartite four kinds of sample data. The audio signal of 24 bits is stored as it is in an audio data memory 24. A maximum peak detection circuit 32 detects a maximum value related to the audio signal of a certain pitch detection retrieval section. Based on the maximum value, the relevant data of the quadripartite sample data of eight bits stored in the memory 26 are read out (it is called as bit shift processing), and self correlation is calculated in a pitch detection circuit 28, and the pitch is detected. When the pitch is detected, the audio signal stored in the memory 24 is read out based on the pitch. The pitch correction processing is performed by the read-out control.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an audio signal processing method and apparatus, and more particularly to a pitch correction processing method and apparatus.

[0002]

2. Description of the Prior Art Digital audio signals are typically 2
Since it is quantized by 4 bits or 16 bits, the signal processing is also performed in the bit length matching the quantized bit even in the pitch correction processing.

[0003]

The signal processing with a bit length of 24 bits or 16 bits complicates the signal processing circuit, requires a long calculation time, and is expensive.

An object of the present invention is not to depend on the quantized bit length of an audio signal, but the pitch width can be accurately detected with the bit length as short as possible, and thus the pitch correction processing can be performed accurately. And to provide the device.

Another object of the present invention is to provide an audio signal processing method and apparatus capable of accurately detecting the pitch width of a stereo signal and accurately performing pitch correction processing.

[0006]

According to the present invention, in the audio signal processing method for detecting the pitch of a digital audio signal and performing pitch correction processing using this pitch, each of the digital audio signals is While storing the code, it is divided into a plurality of bits of sample data and held, a maximum value is detected for the plurality of digital audio signals in a pitch detection search section, and based on the detected maximum value, the maximum value is detected. There is provided an audio signal processing method, characterized in that one type of held sample data is selected, and pitch detection in the pitch detection search section is performed using the selected sample data.

Specifically, in the pitch detection, the autocorrelation of the selected sample data is calculated, and the interval showing the maximum intensity is detected as the pitch width.

[0008] Preferably, when the lower data exceeds a predetermined value, it is limited to within the predetermined value.

Further, preferably, when the digital audio signal is a stereo audio signal, the bit shift processing is performed after adding the digital audio signals of the related channels as a stereo signal.

More preferably, when the data indicating the pitch component of the audio signal used for pitch detection is lost, pitch correction processing is performed using the pitch width obtained in the previous pitch detection search section.

According to the present invention, there is also provided an apparatus for implementing the above audio signal processing method. That is, the audio signal processing device of the present invention which performs the pitch correction processing divides the input audio signal signal into a predetermined bit while holding the first memory means for directly storing the input audio signal and the input audio signal signal. Based on a second memory means for holding a plurality of sample data, a maximum value detection stage for detecting the maximum value of the audio signal in a certain pitch detection search section, and a maximum value detected by the maximum value detection means. , A sample data reading means for reading appropriate sample data of a plurality of sample data for the one audio signal from the second memory means, and a pitch width is detected by using the read sample data. Pitch calculation means, first memory means, second memory means, maximum value detection means, sample data Heading and a control means for controlling the means and pitch calculation means. Further, it further comprises pitch correction processing means for reading the audio signal stored in the first memory means based on the detected pitch width.

[0012]

The circuit is simplified because the bit shift processing is performed to extract the signal component (sample data) of the audio signal necessary for pitch detection, and the pitch width is detected using this short bit length sample data. Therefore, the calculation time can be shortened.

In the bit shift, when the sample data exceeds a predetermined value, the sample data is limited within the predetermined value and can be used for pitch detection.

When the digital audio signal is a stereo audio signal, the digital audio signals of related channels are added as a stereo signal to enable pitch detection common to both of the two channels of audio signals.

When the data indicating the pitch component of the audio signal used for pitch detection is lost, the pitch correction processing is continued using the pitch width obtained in the previous pitch detection search section.

[0016]

1 is a block diagram of a program play system to which an audio signal processing method and apparatus of the present invention are applied. Program play refers to special reproduction processing in which, when reproducing a video signal (video signal or image signal) and an audio signal (audio signal), the reproduction speed is changed to shorten or extend the reproduction time. In FIG.
A block A (upper block) shows a part operable at a variable speed, and a block B shows a part operable at a normal speed. In the program play system, a video recording / reproducing apparatus (VTR apparatus, not shown) reproduces a video signal and an audio signal from a video tape 2 on which a video signal and an audio signal (collectively referred to as an AV signal) are recorded. . During this regeneration stage,
Error correction processing is performed in the error correction circuit (ECC circuit) 4 in the VTR device. The error-corrected video signal is applied to the frame synchronizer 6, and the error-corrected audio signal is output to the pitch correction processing circuit 8.
Is applied to The frame synchronizer 6 matches the reproduction frequency and the external frequency, performs frame dropping or addition of images, and outputs a processed image signal. The pitch correction processing circuit 8 and the memory 10 control the data amount of the variably reproduced audio signal per unit time, and change the reproduction speed without changing the frequency of the audio signal.

The pitch correction process will be described in detail with reference to FIG. FIG. 2A is a waveform diagram of the original audio signal. Each audio signal is composed of a fundamental wave called a peculiar pitch, and the frequency of voice is determined by a large swell of pitches. That is,
It is determined by the interval of the peaks of the waveform shown in FIG. Therefore, if data is deleted or added in fundamental wave units, the amount of data can be controlled without affecting the overall frequency. FIG. 2B shows the waveform of the audio signal subjected to the pitch correction processing. In this example, FIG.
The audio signal of the last pitch shown in (A) is deleted. Although the audio signal from the ECC circuit 4 is variably reproduced, the pitch correction processing circuit 8 temporarily records the audio signal in the memory 10 and then, when the audio signal is read from the memory 10 again, the pitch unit data is recorded. Jump over and read. As a result, FIG. 2 (B)
The pitch-corrected audio signal shown in is obtained.

In the pitch correction processing in the pitch correction processing circuit 8, it is important to detect the pitch (pitch width). To detect the pitch width, the pitch correction processing circuit 8 obtains the autocorrelation of the audio signal recorded in the memory 10. The pitch detecting method will be described with reference to FIGS. 3 and 4. The pitch correction processing circuit 8 is an audio signal recorded in the memory 10 in a certain search section, which is 1 in this example.
For 024 audio signals, two audio signals (samples) are read out at the same time and their autocorrelation is calculated. That is, when the start point of a certain search section is determined, the pitch correction processing circuit 8 causes the memory 10
From that point, 512 samples are read from the memory 10 as a reference data string, at the same time, data of the same reference data string, 512 samples are read, and two samples in a corresponding order are multiplied, The results of these multiplications are added. The result of calculating this for 512 samples is the lag 0, that is, the autocorrelation result in the state where there is no delay in the data strings of the two systems. Then
The 512 pieces of data of the reference data row and the 512 pieces of data at a position 1 to a few points away from the start point of the reference data row are read at the same time, and the data in the corresponding order are multiplied by each other. Add the results. The result of calculating this with respect to 512 samples is the lag 1, that is, the autocorrelation result in a state in which the data strings of the two systems are delayed by a certain unit amount. Hereafter, similarly,
The autocorrelation of lag 2 to lag 114 is calculated. When the above results are plotted, the graph illustrated in FIG. 4 is obtained.
The pitch width is the interval where the autocorrelation is strong, that is, the peak portion in FIG. The pitch correction processing circuit 8 detects such peaks and determines the pitch width.

In digital audio processing, 16-bit or 24-bit quantized data is generally used. The pitch correction processing circuit 8 detects the pitch width, for example, in the case of a 16-bit audio signal, calculates the autocorrelation of one lag,
The multiplication of (16 bits × 16 bits) and the addition process of adding the multiplication results are performed 512 times. The operation result is normally stored in 16 bits. In the case of a 24-bit audio signal, the pitch correction processing circuit 8
Performs (24 bits × 24 bits) multiplications and additions 512 times in order to calculate one-lag autocorrelation. The operation result is 24 bits. Usually, the above-mentioned calculation for pitch width detection is performed according to the number of bits of quantization.
That is, in the case of 24-bit quantized data, the above operation of 24 bits is performed, and in the case of quantized data of 16 bits, the above operation of 16 bits is performed. However, if this is left as it is, the number of bits is large, so that the arithmetic circuit becomes large-scaled. Whether the arithmetic circuit is a digital signal processor (DSP) or an IC circuit, an arithmetic operation with a long bit number complicates the circuit configuration. The price will also increase. In addition, it takes calculation time.

Therefore, it is possible to reduce the number of bits. For example, as shown in FIG. 5, when the dynamic range is 16 bits, pitch width detection can be performed by using 16-bit samples as 8-bit samples. The reason is that even if the dynamic range is 16 bits, even with half the number of bits, the waveform of the mountain portion can be sufficiently expressed. Similarly, in the case of the digital audio signal shown in FIG. 5, a sample having a dynamic range of 24 bits can be used as a 12-bit sample for pitch width detection data. However, as shown in FIG.
Maximum amplitude is 1 for the dynamic range of bits
If it is less than half of 6 bits, it is meaningless to calculate the autocorrelation by reducing the number of bits to 8 bits, and the pitch width cannot be detected. Due to the above circumstances, the pitch correction process has been performed so far by using the sample of the number of bits matching the number of quantization bits. As a result, the configuration of the arithmetic processing circuit becomes complicated,
The price was high and the calculation time was long.

A method for solving the above problem will be described. FIG. 7 is a block diagram of the audio signal processing device 20. The audio signal processing device 20 includes a data control circuit 22, an audio data memory 24, a pitch calculation data memory 2
6, pitch detection circuit 28, memory address counter 3
It has 0 and the maximum peak detection circuit 32. In the present embodiment, a case will be described in which an audio signal having a dynamic range of 24 bits is used as a sample of 8 bits for pitch correction processing, in other words, for pitch width detection.

A 24-bit reproduced audio signal reproduced from a recording medium such as the video tape 2 shown in FIG. 1 and subjected to error correction in the ECC circuit 4 is applied to the data control circuit 22. The data control circuit 22 stores the 24-bit reproduced audio signal S22A as it is in the audio data memory 24. The audio data memory 24 stores and holds the reproduced audio signal S22A as it is. Audio data memory 24
When the audio signal is read again from, the reading order is controlled by the address given from the memory address counter 30, so that the read signal S24 is the input reproduced audio signal S.
22A (for example, for FIG. 2A), the pitch-corrected signal is obtained (for example, the signal shown in FIG. 2B). In the audio data memory 24, a 24-bit reproduced audio signal is divided into 4 parts in a unit of code + 7 bits of data, totaling 8 bits, and stored as 4 pieces of partial data of 1 sample. The details will be described later. Data control circuit 22
Stores the audio signal used for pitch width detection in the pitch detection circuit 28 in the pitch calculation data memory 26. However, the data control circuit 22 causes the pitch calculation data memory 26 to store four types of bit-shifted audio signals for one audio signal sample, as described above. The data stored in the pitch calculation data memory 26 is the data obtained by adding the two-channel stereo signals, and is the data in which 24 bits are changed to 8 bits. Does not increase in quantity.

The bit shift in the data control circuit 22 will be described. It will be described how an audio signal having a dynamic range of 24 bits can be detected as an 8-bit sample for pitch detection. Since an audio signal includes strength and weakness, some audio signals have a maximum amplitude over a dynamic range of 24 bits as shown in FIG. 5, and some audio signals have a maximum amplitude of less than 8 bits as shown in FIG. Also exists. Thus, pitch detection cannot be performed even if an audio signal having a maximum amplitude of less than 8 bits is used as it is. In the present invention, FIG.
For an audio signal with a small amplitude illustrated in (A), for the audio signal in the central portion of FIG.
Bit shift (amplitude amplification) is performed as illustrated in FIG. That is, the lower bits of the audio signal illustrated in FIG. 8A are used. However, when performing the bit shift, it is fixed to a uniform shift amount within the pitch search section. Since the autocorrelation is calculated by using the bit-shifted samples of a plurality of audio signals in a certain search section, it is sufficient that the same section has the same dynamic range. At that time, if there is a data sample larger than the shift amount, the underflowed portion, U
F1, UF2 or overflowed parts OF1, O
A limiter is applied to F2 to clip it. In practice, the sign (sign) of the audio signal stored in the pitch calculation data memory 26 is stored, and the lower 7 bits are extracted and extracted as an 8-bit sample. At this time, samples exceeding 7 bits are samples within 7 bits. It is generally considered that a plurality of audio signals in the same pitch search section have almost the same amplitude. However,
There are variations in their amplitude. Therefore, as described above, four types of bit-shifted sample data obtained by dividing one 24-bit audio signal into four are stored in the pitch calculation data memory 26, and the maximum peak detection circuit 32 detects the maximum peak. And use the appropriate bit-shifted samples. The details will be described later.

The reproduced audio signal S22B is applied to the maximum peak detection circuit 32 after the data of the two channels of the stereo pair have been added. Maximum peak detection circuit 32
Detects the maximum peak of the audio signal in the pitch detection search section. The maximum peak detected by the maximum peak detection circuit 32 is applied to the memory address counter 30 and used as a memory address signal for selecting the optimum sample of the four kinds of samples in the pitch operation data memory 26. . Pitch detection circuit 28
Performs a calculation for pitch detection using the data S26 read from the pitch calculation data memory 26 based on the memory address from the memory address counter 30. As the calculation for this pitch detection, the above-described autocorrelation calculation is preferably performed. However, other calculation methods for detecting the strength of the audio signal, such as a method for calculating the power spectrum, can also be applied. However, such operations can be pushed side by side, for example,
Since multiplication and addition are repeated using about 512 sample data for about 1024 data sections as the pitch detection search section, not limited to autocorrelation,
The calculation is enormous. Therefore, the short bit length of the sample data used for pitch detection is very advantageous for calculation.

When the pitch width is detected, the data control circuit 22 detects the audio data, based on the pitch width.
The read address of the sample data read from the memory 24 is controlled to perform the pitch correction processing illustrated in FIG.

As described above, even for an audio signal having a quantized length of 24 bits, the pitch width can be detected by using an audio signal having a bit length of 8 bits. Since the autocorrelation result for audio signals in the same pitch search section only needs to know the relative value, even if the bit length of the audio signal is shortened in pitch width detection, the accuracy of pitch width detection is not affected. . Therefore, the pitch correction process can be performed accurately. Since pitch width detection processing is performed using an audio signal with a short bit length,
The circuit configuration becomes simple and the calculation time can be shortened. Further, the cost of the circuit that performs the pitch correction process is reduced.

Next, the processing for a stereo signal will be described. FIG. 10 shows the processing circuit configuration. The stereo signal pitch detection circuit 40 shown in FIG.
It includes a parallel (S / P) conversion circuit 44, an adder 46, a data control circuit 42 having a bit shifter 48, a pitch calculation memory 50, and a pitch calculation unit 52. The stereo signal pitch detection circuit 40 shown in FIG.
Related to the data control circuit 22, the pitch calculation data memory 26, and the pitch detection circuit 28 shown in FIG.
The maximum peak detection circuit 32, the memory address counter 30, and the audio data memory 24 shown in FIG. 7 are added to the stereo signal pitch detection circuit 40 shown in FIG. 10 to form a circuit for detecting the pitch width.

The reproduced audio signal input in FIGS. 7 and 10 usually has four channels of 24-bit data. Usually, a combination of channel 1 and channel 2 and a combination of channel 3 and channel 4 form a stereo signal, so pitch detection is also performed in consideration of this. In pitch detection for a stereo signal, it is important to detect a pitch common to each signal. Therefore, the signals of the two channels are added as preprocessing for pitch detection. As the preprocessing for this pitch detection, the reason why addition is performed before bit shifting will be described. If bit shift is performed, the result of clipping the maximum value of the sample data may be added. In this case, there is a possibility that the pitch characteristics of the data may be lost by the addition. Furthermore, when the stereo signals are added, if the signals of the two channels have opposite phases, the addition result will lose the amplitude component at all. Therefore, the audio signals of two channels are added before the bit shift.

The S / P conversion circuit 44 converts a 24-bit reproduced audio signal input in serial format into 24-channel stereo audio signals of two channels. The adder 46 adds the two-channel stereo audio signals before the bit shift in the bit shifter 48. The bit shifter 48 performs the bit shift processing illustrated in FIGS. 8 and 9 on the stereo audio signal in which the audio signals of 2 channels are added. In this way, the pitch calculator 52 uses the bit-shifted audio signal as necessary,
Performs pitch detection calculation. As described above, according to the stereo signal pitch detection circuit 40 of the second embodiment, since the two channels are added in the adder 46 before the processing of the bit shifter 48, after the bit shift is performed. The accuracy of pitch detection is improved and the reliability of pitch correction processing is also improved as compared with the case where stereo addition processing is performed.

When several kinds of bit shifts are set for an audio signal, from several kinds of sample data,
Strength (amplitude) of sample data in the pitch detection search section
A method of selecting one according to the above will be described. If the value of the bit shift is fixed to the maximum, when the reproduced audio signal has the maximum amplitude, the limiter of the maximum value may be always applied as shown in FIG. The pitch width cannot be detected from the data subjected to the limiter as shown in FIG. Therefore, in the present invention, as described below, the bit shift amount is adaptively adjusted according to the size of the input audio signal.

Details of the adaptive bit shift processing will be described below. When cutting out 8-bit data from a 24-bit audio signal, it is desirable to bit-shift to the maximum value in the range where the limiter is not applied to the data in order to save the original audio signal. In the present embodiment, the 24-bit audio signal shown in FIG. 12 is divided into four types of 8-bit data and stored in the pitch calculation memory 50, as shown in FIG.
Each of the 8-bit data stores the code of the original audio signal. Therefore, the effective bit length is 7 bits. In order to maintain the continuity of the four types of data, adjacent data includes overlapping bit data. 12
The pitch detection search section 1 and the pitch detection search section 2 each include about 1024 audio signals each having one peak. These autocorrelation operations are a fairly large amount of operations and are time consuming. The pitch detection is performed after all the audio signals are stored in the pitch calculation data memory 26 shown in FIG. 7 or the pitch calculation memory 50 shown in FIG. Therefore, the pitch calculation data memory 26 or the pitch calculation memory 5
When the audio signal is stored in 0, the maximum value of the audio signal in a certain pitch detection search section is detected. The maximum value is detected by the maximum peak detection circuit 32 in the audio signal processing device 20 shown in FIG. In the stereo signal pitch detection circuit 40 shown in FIG. 10, the maximum value is detected after the stereo signal addition processing is performed in the adder 46. In the audio signal processing device 20 of FIG. 7, the maximum peak detection circuit 32 stores the maximum value detected in a certain pitch detection search section, and the maximum value is used to control the address of the memory address counter 30, 1
As shown in FIG. 3, four kinds of 8-bit sample data, which are stored in the pitch calculation data memory 26, are read out for each 24-bit audio signal.

When the audio signal is a stereo signal, the 4-division sample data shown in FIG. 13 is stored in the pitch calculation memory 50 in the state shown in Table 1.

[0033]

[Table 1]

In Table 1, the X system is a stereo signal obtained by adding the audio signal of channel 1 and the audio signal of channel 2 each having a length of 24 bits, and the bit length of the added stereo audio signal is 24 bits, The first sample data is [sign, 22 to 16 bits], the second sample data is [sign, 16 to 10 bits], the third sample data is [sign, 11 to 5 bits], and the fourth sample data. Is [sign, 6-0 bits]
Is divided into four and stored in the pitch calculation memory 50.
The Y system is a stereo signal in which an audio signal of channel 3 and an audio signal of channel 4 each having a length of 24 bits are added, the bit length of the added stereo audio signal is 24 bits, and the first sample data is [ Code, 22 to 16 bits], the second sample data is [code, 16 to 10 bits], the third sample data is [code, 11 to 5 bits], and the fourth sample data is [code, 6 to 0]. Bit] and stored in the pitch calculation memory 50.

The 8-bit sample data read as described above from the pitch calculation memory 50 is applied to the pitch detection circuit 28 and used for pitch detection.

In the waveform diagram of FIG. 12, the audio signal in the pitch detection search section 1 has the maximum amplitude in the full 24-bit dynamic range, and the other continuous waveform portions with small amplitude are within 8 bits. Therefore, the bit detection is not substantially performed, and the pitch detection is performed using the most significant 8 bits of sample data. 12
The maximum value of the audio signal in the pitch detection search section 2 is small. In this case, pitch detection is performed using the bit-shifted least significant 8 bits of sample data.

A third embodiment of the present invention will be described. When pitch detection is performed by performing the bit reduction processing described above and pitch correction processing is performed, there are cases where pitch detection cannot be performed depending on the state of the audio signal. For example, if it is not a stereo signal (in the case of a monotone audio signal), pitch detection cannot be performed if the original audio signal itself has no pitch component.
Also, for example, as illustrated in FIG.
When the audio signal of FIG. 10 is inverted on the way, the stereo signal pitch detection circuit 4 of FIG.
When the audio signal of channel 1 and the audio signal of channel 2 are added by the adder 46 of 0, the opposite phases cancel each other out, and the audio signal component disappears from the middle as shown in FIG. With such an audio signal, in the pitch detection search section 2,
The pitch cannot be detected. The third embodiment relieves such a case.

As shown in FIG. 15, even if there is an audio signal whose pitch cannot be detected on the way, in consideration of the continuity of the audio signal, the continuity with the audio signal of the preceding pitch detection search section is maintained. Therefore, in such a case, the pitch width of the previous pitch detection search section that is normally detected is used. Therefore, in place of the pitch calculator 52 in the stereo signal pitch detection circuit 40 of FIG. 10, a pitch calculator holding circuit 60 including a pitch calculator 62 and a pitch information holding circuit 64 shown in FIG.
To provide. When the pitch is normally detected in the pitch calculator 62, the pitch information holding circuit 64 is switched to the contact a side, and the pitch width information is stored in the pitch calculation holding circuit 6.
Output from 0. If the pitch calculator 62 cannot detect the pitch normally, the pitch information holding circuit 6
4 is switched to the contact b side, and the pitch width information calculated in the previous pitch detection search section is output from the pitch calculation holding circuit 60 via the pitch information holding circuit 64.
That is, the pitch width information obtained in the previous pitch detection search section is output. As a result, the pitch correction process can be continued even if the pitch component does not exist instantaneously.

The pitch calculation holding circuit 60 shown in FIG.
Can also be applied to the pitch detection circuit 28 shown in FIG.

The embodiments of the present invention are not limited to those described above, and various other structures can be adopted. For example, the embodiments described above can be applied independently, and they can be arbitrarily combined.

In the above embodiment, if the original digital audio signal is 24 bits, 8
Although the example of performing the bit shift to the bit sample data is shown, the relationship of the bit shift can be arbitrarily set. For example, a 24-bit quantized audio signal
It is also possible to perform bit shift as sample data of 10 bits in total including sign + 9 bits. In this case, for example, the first sample data = [code + 22 to 14-bit data], the second sample data = [code + 12 to 4]
Bit data], third sample data = [sign + 8]
~ 0-bit data]. Alternatively,
In order to further reduce the number of bits, of course, a 24-bit quantized audio signal can be bit-shifted as 6-bit (sign + 5 bit) sample data and divided into 5 sample data per audio signal. Of course, the result of bit-shifting the 16-bit quantized audio signal to 8 bits, 6 bits, or the like can also be used.

Although the present invention has been described in connection with the program play system shown in FIG. 1, the present invention is not limited to the program play system.

[0043]

According to the present invention, since bit shifting is performed, pitch detection can be performed with a small number of bits even for a digital audio signal having a small amplitude.

Further, in the present invention, since the two channels of audio signals are added to the stereo digital audio signal before the bit shift, the accuracy of pitch detection becomes high.

Further, in the present invention, when the pitch component of the audio signal disappears instantaneously, the pitch width obtained in the previous pitch detection search section is used in consideration of the continuity of the audio signal. Even in the presence of such a defective audio signal, the pitch correction process can be relieved.

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration of a program play system as an example to which an audio signal processing method and apparatus of the present invention are applied.

2 is a graph showing pitch correction processing, FIG. 2 (A) is a waveform diagram of an original audio signal before pitch correction processing, and FIG. 2 (B) is an audio signal after pitch correction processing. It is a waveform diagram.

FIG. 3 is a graph showing how an autocorrelation of an audio signal is calculated in order to detect a pitch width for performing pitch correction processing.

FIG. 4 is a graph showing detection of a pitch width for performing a pitch correction process.

FIG. 5 is a waveform diagram of a digital audio signal which is a target of pitch correction processing.

FIG. 6 is a waveform diagram of a digital audio signal that is a target of pitch correction processing, and is a waveform diagram when the maximum amplitude is small with respect to the dynamic range.

FIG. 7 shows a first audio signal processing apparatus according to the present invention.
It is a block diagram of an Example.

8 is a diagram illustrating a bit shift process in a pitch correction process, FIG. 8 (A) is a waveform diagram of a digital audio signal when the amplitude is small, and FIG. 8 (B) is a bit shift process. It is a wave form diagram which shows the result.

FIG. 9 is a graph showing a result of bit-shifting a 24-bit digital audio signal into an 8-bit digital audio signal (sample data) in the audio signal processing device of the present invention.

FIG. 10 is a configuration diagram of a second embodiment of an audio signal processing device of the present invention.

FIG. 11 is a waveform diagram of a digital audio signal that is a target of pitch correction processing, and is a waveform diagram when a limiter is frequently applied to bit-shifted data.

FIG. 12 is a graph illustrating a bit shift for adaptively detecting pitch.

FIG. 13 is a graph showing the bit shift processing illustrated in FIG.

FIG. 14 is a graph showing processing when pitch detection is not possible.

FIG. 15 is a graph showing a rescue processing result when the pitch of the digital audio signal shown in FIG. 14 cannot be detected.

FIG. 16 is a configuration diagram of a pitch width processing circuit as a third embodiment in the audio signal processing device of the invention.

[Explanation of symbols]

 2 ... Video tape 4 ... Error correction circuit 6 ... Frame synchronizer 8 ... Pitch correction processing circuit 10 ... Memory 20 ... Audio signal processing device 22 ... Data control circuit 24 ... Audio data memory 26 ... -Pitch calculation data memory 28-Pitch detection circuit 30-Memory address counter 32-Maximum peak detection circuit 40-Stereo signal pitch detection circuit 42-Data control circuit 44-S / P conversion circuit 46- Adder 48-bit shifter 50-pitch calculation memory 52-pitch calculator 60-pitch calculation holding circuit 62-pitch calculator 64-pitch information holding circuit

Claims (11)

[Claims] 1. An audio signal processing method for detecting a pitch of a digital audio signal and performing a pitch correction process using the pitch, wherein each of the digital audio signals has a plurality of bits while preserving a code. The sample data is divided and held, the maximum value is detected for the plurality of digital audio signals in a certain pitch detection search section, and one of the held sample data is detected based on the detected maximum value. An audio signal processing method, comprising: selecting, and using the selected sample data, performing pitch detection in the pitch detection search section. 2. The pitch detection calculates an autocorrelation of the selected sample data, and detects an interval showing the maximum intensity as a pitch width.
The described audio signal processing method. 3. The audio signal processing method according to claim 1, wherein, when the lower data exceeds a predetermined value, the lower data is limited to the predetermined value. 4. The audio signal according to claim 1, wherein when the digital audio signal is a stereo audio signal, the bit shift processing is performed after adding the digital audio signals of related channels as a stereo signal. Processing method. 5. When the data indicating the pitch component of the audio signal used for pitch detection is lost, pitch correction processing is performed using the pitch width obtained in the previous pitch detection search section. Any one of the audio signal processing methods. 6. An audio signal processing device for performing pitch correction processing of a digital audio signal, wherein a first memory means for storing an input audio signal as it is and a predetermined code while holding the input audio signal signal respectively. Second memory means for holding as a plurality of sample data divided into a plurality of bits, a maximum value detection stage for detecting the maximum value of the audio signal within a certain pitch detection search section, and a maximum value detection means for detecting the maximum value. Based on the maximum value, using appropriate sample data of the plurality of sample data for the one audio signal from the second memory means, using the sample data reading means, and the read sample data, Pitch calculation means for detecting pitch width, first memory means, second memory means The audio signal processing apparatus having the maximum value detecting means, and control means for controlling the means and pitch computing means read sample data. 7. A pitch correction processing means for reading an audio signal stored in said first memory means based on said detected pitch width, further comprising:
The audio signal processing device according to claim 6. 8. The audio signal processing according to claim 6, wherein the pitch calculation means calculates an autocorrelation of the read sample data, and detects an interval showing the maximum intensity of the autocorrelation as a pitch width. apparatus. 9. The audio according to claim 6, wherein the control means and the second memory means limit the read sample data to a predetermined value or less when the read sample data exceeds the predetermined value. Signal processing device. 10. The control means, when the digital audio signal is a stereo audio signal, adds means for adding digital audio signals of channels related as stereo signals, and holds a sign of the addition result. 10. The audio signal processing device according to claim 6, further comprising a bit shift unit that performs a bit shift process as a plurality of sample data divided into predetermined bits and applies the bit shift process to the second memory unit. 11. The pitch calculating means outputs the pitch width obtained in the preceding pitch detection search section when the data indicating the pitch component of the audio signal used for pitch detection is lost. 10. The audio signal processing device according to any one of 10.