JP4332144B2 - Waveform data format discrimination method, signal encoding method, apparatus using the methods, program, and recording medium - Google Patents

Description

  The present invention relates to a method for determining a data storage format of waveform data whose data storage format is unknown, a signal encoding method using the method, an apparatus, a program, and a recording medium using those methods.

There are many proposals for compression coding without distortion. For example, for text data such as sentences and programs, a general-purpose compression encoding method (encoding method for non-acoustic signals) called ZIP is often used. On the other hand, as a compression encoding method without distortion such as a music signal, for example, there is an encoding method (encoding method for an acoustic signal) of Patent Document 1. This method can compress much more effectively than ZIP.
However, in the case of acoustic data such as music and voice, the encoding efficiency is greatly reduced unless encoding is performed according to the storage format of the acoustic data described in the header format. The storage format of the acoustic data includes information such as whether the acoustic data is big endian or little endian, how many bytes of data, integer representation or floating point representation. In other words, for efficient encoding, it is necessary to know all of the storage format of such acoustic data, and when part or all of the storage format is unknown, it is efficient to use the characteristics specific to acoustic data. Encoding cannot be realized.
JP 2005-115267 A

  In the case of the prior art, if part or all of the storage format of the acoustic data in the digital data is not known, the encoding using the characteristics inherent in the acoustic data cannot be performed, so the encoding efficiency is greatly reduced. Become. Thus, even if it is known that the input data is acoustic data, the conventional technology has a problem that it is difficult to handle the input data as acoustic data if the data storage format is unknown. This also applies to digital waveform data obtained by sampling analog waveform data. An object of the present invention is to determine a data storage format when it is known that the input digital data is waveform data such as acoustic data but the storage format of the waveform data is unknown.

In the present invention, although it is known that it is waveform data, when part or all of the waveform data storage format is unknown, the input digital waveform data is converted into each of a plurality of preset data storage formats. The likelihood of waveform data in the case of a sample value sequence is determined using the relationship between sample values in the sample value sequence. The data storage format corresponding to the sample value sequence determined to be the most waveform data is set as the data storage format of the input digital waveform data.
Also, all or part of the input digital data is encoded based on the result of the data storage format discrimination. Furthermore, if the waveform data likelihood as a result of the data storage format determination is within a predetermined range, the waveform signal is encoded in the corresponding data storage format. Is encoded.

  According to the present invention, the data storage format can be determined even when the data storage format of the input digital waveform data is not known. Further, even if it is not actually encoded, it can be determined which waveform data format is efficient with a low calculation amount, and can be encoded efficiently.

  In the following, an embodiment in which the present invention is applied to acoustic data, which is the most representative analog waveform data, will be described. In order to avoid duplication of description, the same number is assigned to the constituent unit having the same function and the processing step for performing the same process, and the description is omitted.

[First Embodiment]
An example of the functional configuration of the acoustic data format discrimination unit of the present invention is shown in FIG. Moreover, the processing flow of the acoustic data format discrimination | determination part 100 is shown in FIG. The acoustic data format determination unit 100 includes a data recording unit 110, an index determination unit 120, and a data storage format determination unit 130. The index determination unit 120 includes a storage format recording unit 121 and an index calculation unit 124.
The input digital acoustic data is first recorded in the data recording unit 110 (S110). The index calculation unit 124 of the index determination unit 120 sequentially selects a data format from a plurality of data formats set in advance in the storage format recording unit 121, and part or all of the input digital acoustic data is respectively by using the relationship between the sample values in the case of that the data format, to evaluate the acoustic data likelihood in the case where the digital audio data inputted as the respective data format (S124). Here, the preset data format is the number of bits per sample of a plurality of types, whether it is an integer representation or a floating-point representation, the byte storage order for a plurality of types of samples, the number of channels of a plurality of types, etc. It is. Digital sound data used for evaluation may be all input or a part thereof. When a part is used, for example, the first thousands of samples to tens of thousands of samples of the digital sound data may be used. The evaluation of the likelihood of acoustic data is performed, for example, every several thousand frames.

ただし、ｐはあらかじめ定めた正の整数、ｋ_ｉ（ｉ＝１〜ｐ）はＰＡＲＣＯＲ係数。
Ｆは、ｐ次の予測の場合の予測残差のエネルギーに対する入力信号のエネルギーの比を近似するものである。Ｆの値が大きい方が、そのデータ形式らしいと言える。従って、入力されたデジタル音響データを音響信号符号化処理により圧縮する場合は、入力されたデジタル音響データを、Ｆの値が大きいデータ形式で圧縮すれば、圧縮後のデータ量を少なくすることができる。つまり、高い圧縮効果が期待できる。なお、例えば、フレーム単位に音響データらしさの指標を計算する場合に、想定したデータ形式でのサンプル値列が複数フレーム分ある場合には、データ形式ごと、フレームごとの複数の指標の値が求まることになる。このように同じデータ形式で複数の指標の値がある場合には、それらの中の最大の値、すなわち、最も音響データらしい指標の値を、当該データ形式のサンプル値列に対する指標の値とする。 Specifically, the following expression is used as an index of the likelihood of sound data.

Here, p is a predetermined positive integer, and k _i (i = 1 to p) is a PARCOR coefficient.
F approximates the ratio of the energy of the input signal to the energy of the prediction residual in the case of p-th order prediction. It can be said that the one with the larger F value seems to be the data format. Therefore, when the input digital sound data is compressed by the sound signal encoding process, if the input digital sound data is compressed in a data format having a large F value, the amount of data after compression can be reduced. it can. That is, a high compression effect can be expected. Note that, for example, when calculating an index of the likelihood of sound data in units of frames, if there are a plurality of sample value sequences in an assumed data format, the values of a plurality of indexes for each data format and each frame are obtained. It will be. Thus, when there are a plurality of index values in the same data format, the maximum value among them, that is, the index value that seems to be the most acoustic data is set as the index value for the sample value sequence of the data format. .

The data storage format determination unit 130 compares the F values corresponding to the respective data formats, and determines the data format having the largest F value as the data format of the input digital acoustic data (S130).
By discriminating the data format in this way, the data format of the input digital acoustic data can be estimated.

ただし、Ｍはあらかじめ定めた正の整数、Ｙ_ｊ（ｊ＝０〜Ｍ−１）はｊ番目の周波数領域係数の２乗値。
Ｅは、パワースペクトルに相当する係数の２乗値の相加平均の比である。係数値がｊによらず一定であればＥ＝１という最低値となる。この場合、入力されたデジタル音響データをそのデータ形式とした場合のサンプル値列は、乱数であり、そのデータ形式では音響データらしいとは言えない。従って、入力されたデジタル音響データを音響信号符号化処理により圧縮する場合に、そのデータ形式でのデジタル音響データとして圧縮しても、圧縮後のデータ量を少なくすることはできない。つまり、圧縮効果は期待できない。一方、係数の２乗値の変動が大きければＥは大きな値となる。周波数領域の係数（スペクトル）の値に大きな変動があるので、そのデータ形式での音響データらしいと言える。従って、入力されたデジタル音響データを音響信号符号化処理により圧縮する場合に、入力されたデジタル音響信号を、そのデータ形式であるとして圧縮すれば、圧縮後のデータ量を少なくできる。つまり高い圧縮効果が期待できる。また、これはサンプル値間の相関が強いことも意味している。
したがって、この方法によっても、入力されたデジタル音響データのデータ形式を推定することができる。 [Modification 1] In the first embodiment, F is used as an index indicating the likelihood of sound data. However, in the case of encoding to be converted into the frequency domain, evaluation can also be performed using the following equation.

Here, M is a predetermined positive integer, and Y _j (j = 0 to M−1) is a square value of the j-th frequency domain coefficient.
E is an arithmetic mean ratio of square values of coefficients corresponding to the power spectrum. If the coefficient value is constant regardless of j, the minimum value is E = 1. In this case, the sample value string when the input digital acoustic data is in the data format is a random number, and it cannot be said that the data format is likely to be acoustic data. Therefore, when the input digital audio data is compressed by the audio signal encoding process, the amount of data after compression cannot be reduced even if the digital audio data is compressed as digital audio data in the data format. In other words, the compression effect cannot be expected. On the other hand, if the variation of the square value of the coefficient is large, E becomes a large value. Since there is a large fluctuation in the value of the frequency domain coefficient (spectrum), it can be said that it seems to be acoustic data in that data format. Therefore, when the input digital sound data is compressed by the sound signal encoding process, the amount of data after compression can be reduced by compressing the input digital sound signal as the data format. That is, a high compression effect can be expected. This also means that the correlation between sample values is strong.
Therefore, also by this method, the data format of the input digital acoustic data can be estimated.

[Modification 2]
Energy / F or energy / E can also be used as an index indicating the likelihood of acoustic data. Energy / F or energy / E corresponds to approximation of the energy of the prediction error, and it can be determined that the smaller the value, the more likely it is acoustic data. Therefore, when this index is used, the data storage format determination unit 130 compares the index (energy / F or energy / E) value corresponding to each data format, and the data format having the smallest value is input. It is discriminated from the data format of the digital sound data (S130).

[Second Embodiment]
FIG. 3 shows a functional configuration example of the encoding apparatus according to the present embodiment. FIG. 4 shows a processing flow of the encoding device 200. The encoding device 200 includes a data recording unit 110, an index determination unit 120, a data storage format determination unit 130, and an encoding unit 240. As can be seen from a comparison between FIG. 3 and FIG. 1, the encoding device 200 in FIG. 3 has a configuration in which an encoding unit 240 is added to the acoustic data format determination unit 100 in FIG. 1.
The processing flow in FIG. 4 is also a processing flow in which step S240 is added to the processing flow in FIG. In step S240, the encoding unit 240 encodes the digital sound data recorded in the data recording unit 110 according to the data format determined by the data storage format determination unit 130. Therefore, the encoding unit 240 needs to include an encoding unit that can support a plurality of data formats. That is, the data format used when the index calculation unit 124 calculates the index (step S124) is limited to the range of encoding methods that can be supported by the encoding unit 240.
Thus, in this embodiment, since the input digital acoustic data can be encoded as having the data format determined to be the most acoustic data, high compression efficiency can be expected.

[Third Embodiment]
In the first embodiment and the second embodiment, it is assumed that the input digital data is acoustic data. The present embodiment also supports a situation where it may not be acoustic data. FIG. 5 shows a functional configuration example of the encoding apparatus according to the present embodiment. FIG. 6 shows a processing flow of the encoding device 300. The encoding apparatus 300 has a configuration in which an encoding method selection unit 350 and an encoding unit 340 are added to the acoustic data format determination unit 100 of FIG. The encoding method selection unit 350 includes an index comparison unit 351. The encoding unit 340 includes an acoustic signal encoding unit 341 and a non-acoustic signal encoding unit 342. The non-acoustic signal encoding unit 342 executes a general-purpose encoding method such as ZIP.

Steps S110 to S130 in the processing flow of FIG. 6 are the same as those in the first embodiment (FIG. 2). After that, the index comparison unit 351 of the encoding method selection unit 350 inputs the acoustic data likelihood of the input digital data when the data format is determined to be the most likely acoustic data (an index indicating the possibility of being acoustic data). ) To select whether the encoding is performed by the encoding unit for the acoustic signal or the encoding unit for the non-acoustic signal (S351). For example, it may be determined by comparing with the threshold value whether or not the sound data is more likely than the threshold value. When F or E shown in the first embodiment is used as an index indicating the likelihood of sound data, it is likely that sound data is larger than a threshold (or greater than or equal to a threshold). As will be described in an experimental example described later, for example, a case where the value of F or E is 1000 or more is determined to be acoustic data. Further, when energy / F or energy / E is used as an index indicating the likelihood of sound data, it is likely that the sound data is smaller than the threshold (or less than the threshold).
If it is determined in step S351 that the data is likely to be acoustic data, the acoustic signal encoding unit 341 encodes the input digital data recorded in the data recording unit 110 in the data format determined by the data storage format determination unit 130. (S241). If it is determined in step S351 that the sound data does not appear, the non-acoustic signal encoding unit 342 encodes the input digital data recorded in the data recording unit 110 (S242).
As described above, in the present embodiment, the encoding method is determined after evaluating the sound data quality of the input digital data. Therefore, even when data other than the sound data is input, a certain degree of compression efficiency is achieved. Can be secured.

[Modification]
In the third embodiment, the determination is simply made by comparison with the threshold value in step S351. However, depending on whether the data format of the input signal is 1 byte or 2 bytes or more, or depending on the magnitude of energy when the sample value sequence is the amplitude of the waveform, it is better to use the encoding means for sound. In some cases, it may be better to use acoustic encoding means. Therefore, in this modification, an example in which the processing of the encoding method selection unit 350 is changed is shown.

  FIG. 7 shows a processing flow of the encoding apparatus 300 according to this modification. In this processing flow, steps S352 to S356 are executed instead of step S351 in FIG. Steps S352 to S356 will be described below. The encoding method selection unit 350 confirms whether the determined data format of the input digital data is 1 byte (S352). When step S352 is No, the encoding method selection unit 350 calculates energy when the sample value sequence is the waveform amplitude (S353), and checks whether the energy is smaller than the threshold (S354). As the threshold value, for example, there is a method of setting the energy to 1/100 of the maximum amplitude. When step S354 is Yes, it progresses to step S241. When step S354 is No, the index comparison unit 351 compares whether the index indicating the likelihood of acoustic data is more likely to be acoustic data than the threshold (S355). In this step, the threshold value for F or E, which is an index indicating the likelihood of acoustic data, may be about 100. When step S355 is Yes, it progresses to step S241, and when it is No, it progresses to step S242. If step S352 is Yes, the index comparison unit 351 compares whether the index indicating the likelihood of acoustic data is more likely to be acoustic data than the threshold (S356). This step is the same as step S351 in the third embodiment (FIG. 6).

Thus, since the encoding method is selected according to the data format, higher compression efficiency can be expected than in the third embodiment.
In the first to third embodiments, the case where the input data is digital acoustic data obtained by sampling an acoustic signal has been described. However, if the input data is obtained by sampling analog waveform data, the present invention can be used even with other time series data. In this case, the present invention can be applied to general waveform data only by changing “determination of the likelihood of acoustic data” in each embodiment to “determination of the likelihood of waveform data”.
In addition, each said embodiment can also read and carry out the program which makes a computer perform each step of the said method. Also, as a method for reading into the computer, the program is recorded on a computer-readable recording medium, and the program is read from the recording medium into the computer, or the program recorded in the server or the like is read into the computer through an electric communication line or the like. There are methods.

[Experimental example]
In order to evaluate the effects of the present invention, simulation results in the second embodiment are shown below. The simulation conditions are as follows. The input signal is waveform data of an image having a data amount of 1 byte / sample and 2048 samples / frame (144054 bytes: about 70 frames). The data format set in advance is the number of channels (1, 2, 3). The calculated index is energy, F (p = 1, 2, 3, 20). When this entire data (about 70 frames) is actually compressed, the compression efficiency is highest when the number of channels is assumed to be 3 and the data is assumed to be 29.41%, and then the number of channels is assumed to be 1 Is 36.13%, and 40.00% when the number of channels is assumed to be 2. In the simulation, energy and F were calculated for the first nine frames of the input signal. In other words, the number of frames for one channel is 9 frames when the number of channels is 1, 5 frames of the first channel and 4 frames of the second channel when the number of channels is 2, and when the number of channels is 3. This is three frames for each channel.

FIG. 8 shows the results when the data format is assumed to be 1 with the number of channels, FIG. 9 shows the results when the data format is assumed with 2 channels, and FIG. 10 shows the results when the number of channels is assumed to be 3. Since F is a large value regardless of the value of p, it can be seen that it is desirable to compress in the data format with 3 channels. It can also be seen that it is desirable to use a data format with 3 channels, depending on energy / F.
Note that the value of F increases monotonically as p is increased. Further, when the value of E is calculated as a power spectrum obtained by smoothing Y _j or a power spectrum obtained by adding a small positive value, it is known that the value approximates the value of F when p is increased. . Therefore, although it is understood that the number of channels should be 3 even when p is small, it is desirable to use F or E or energy / F or energy / E with p as a value of about 20 as an index. Moreover, at least p should be 3 or more.
As described above, it is understood that the value of the number of channels can be predicted by obtaining F.

The figure which shows the function structural example of the acoustic data format discrimination | determination part of 1st Embodiment. The figure which shows the processing flow of the acoustic data format discrimination | determination part of 1st Embodiment. The figure which shows the function structural example of the encoding apparatus of 2nd Embodiment. The figure which shows the processing flow of the encoding apparatus of 2nd Embodiment. The figure which shows the function structural example of the encoding apparatus of 3rd Embodiment. The figure which shows the processing flow of the encoding apparatus of 3rd Embodiment. The figure which shows the processing flow of the encoding apparatus of the modification of 3rd Embodiment. The figure which shows the simulation result at the time of assuming that the number of channels is 1 data format. The figure which shows the simulation result at the time of assuming that the number of channels is 2 data format. The figure which shows the simulation result at the time of assuming that the number of channels is 3 data formats.

Claims (18)

Data storage format of the input digital sound waveform data has been found a waveform data type determination method of determining which one of a plurality of data storage format which is set in advance,
For a plurality of preset data formats, the input digital waveform acoustic data is converted using the relationship between sample values when some or all of the input digital acoustic waveform data is in the respective data format. An index determination step for evaluating the likelihood of acoustic data when each data format is assumed;
A waveform data format discrimination method comprising: a data storage format discrimination step for setting a data storage format determined to be most acoustic data as a data storage format of input digital acoustic waveform data. A waveform data format determination method according to claim 1, comprising:
The preset data storage format is the number of bits per sample of a plurality of types, whether it is an integer representation or a floating point representation, the byte storage order for each of a plurality of types of samples, the number of channels of a plurality of types, A waveform data format discrimination method characterized by being any two or more of them. A waveform data format discrimination method according to claim 1 or 2,
As an index of the likelihood of waveform data in the index determination step,

Where p is a predetermined positive integer, k _i (i = 1 to p) is a PARCOR coefficient,
Waveform data format discrimination method characterized by using A waveform data format discrimination method according to claim 1 or 2,
As an index of the likelihood of waveform data in the index determination step,

Where M is a predetermined positive integer, Y _j (j = 0 to M−1) is the square value of the j-th coefficient,
Waveform data format discrimination method characterized by using A signal encoding method using the waveform data format determination method according to any one of claims 1 to 4,
A signal encoding method comprising: an encoding step for encoding all or part of input digital data based on the result of the data storage format determination step. A signal encoding method using the waveform data format determination method according to any one of claims 1 to 4,
A selection step of selecting an encoding method based on the result of the data storage format determination step and the waveform data likeness in the determined waveform data format;
A signal encoding method comprising: an encoding step for encoding all or part of digital data input by the encoding method selected in the selection step. A signal encoding method according to claim 6, comprising:
If the waveform data likelihood in the determined waveform data format is in a predetermined range, the selection step selects an encoding method corresponding to the result of the data storage format determination step, and the determined waveform data format If the waveform data likelihood is outside the predetermined range, the encoding method for the non-waveform signal is selected.
A signal encoding method characterized by the above. A signal encoding method using the waveform data format discrimination method according to claim 3 or 4,
If the value of the index of waveform data likelihood in the determined waveform data format is 1000 or more when calculated in units of 1 byte / sample, and 100 or more when calculated in other units, the data storage format determination step A selection step of selecting an encoding method corresponding to the result, and selecting an encoding method for a non-waveform signal when the value of the index of the likelihood of waveform data in the determined waveform data format is outside the above range; ,
A signal encoding method comprising: an encoding step for encoding all or part of digital data input by the encoding method selected in the selection step. Data storage format of the input digital sound waveform data has been found a waveform data type determination apparatus for determining which of in a preset plurality of data storage format,
A data recording unit for recording the input digital acoustic waveform data;
For a plurality of preset data formats, the input digital waveform acoustic data is converted using the relationship between sample values when some or all of the input digital acoustic waveform data is in the respective data format. An index determination unit that evaluates the likelihood of sound data in the case of each data format ;
A waveform data format discriminating apparatus comprising: a data storage format discriminating unit that sets a data storage format determined to be the most acoustic data as a data storage format of input digital acoustic waveform data. The waveform data format discrimination device according to claim 9,
The preset data storage format is the number of bits per sample of a plurality of types, whether it is an integer representation or a floating point representation, the byte storage order for each of a plurality of types of samples, the number of channels of a plurality of types, A waveform data format discriminating apparatus characterized by being any two or more of them. The waveform data format discrimination device according to claim 9 or 10,
As an index of the likelihood of waveform data in the index determination unit,

Where p is a predetermined positive integer, k _i (i = 1 to p) is a PARCOR coefficient,
A waveform data format discriminating apparatus characterized by using the above. The waveform data format discrimination device according to claim 9 or 10,
As an index of the likelihood of waveform data in the index determination unit,

Where M is a predetermined positive integer, Y _j (j = 0 to M−1) is the square value of the j-th coefficient,
A waveform data format discriminating apparatus characterized by using the above. A signal encoding device comprising the waveform data format discrimination device according to any one of claims 9 to 12,
A signal encoding device comprising: an encoding unit that encodes all or part of input digital data based on a result of the data storage format determination unit. A signal encoding device comprising the waveform data format discrimination device according to any one of claims 9 to 12,
A selection unit for selecting an encoding means based on the result of the data storage format determination unit and the waveform data likeness in the determined waveform data format;
A signal encoding apparatus comprising: an encoding unit that encodes all or part of digital data input by the encoding unit selected by the selection unit. The signal encoding device according to claim 14, wherein
When the waveform data likelihood in the determined waveform data format is within a predetermined range, the selection unit selects an encoding unit corresponding to the result of the data storage format determination unit, and uses the determined waveform data format. If the waveform data likelihood is outside the predetermined range, the encoding means for the non-waveform signal is selected.
A signal encoding device. A signal encoding device comprising the waveform data format discrimination device according to claim 11 or 12,
If the value of the index of waveform data likelihood in the determined waveform data format is 1000 or more when calculated in units of 1 byte / sample, and 100 or more when calculated in other units, the data storage format determination step A selection unit that selects an encoding unit corresponding to the result, and selects an encoding unit for a non-waveform signal when the value of the index of waveform data likelihood in the determined waveform data format is outside the range; ,
A signal encoding apparatus comprising: an encoding unit that encodes all or part of digital data input by the encoding unit selected by the selection unit. Program for executing a computer a method according to any one of claims 1 to 8.   A computer-readable recording medium on which the program according to claim 17 is recorded.

Images