JPH04143970A - Interpolating device for audio signal - Google Patents

Abstract

PURPOSE:To obtain a satisfactory reproducing voice by extracting the front and the rear blocks of a defective block, inverting a time series of an audio signal and weighting it, and thereafter, adding it and executing an interpolation processing, and substituting it for an output signal of the defective block. CONSTITUTION:PCM information DA0 passes through a 1-block delay circuit 14 from a terminal 11 and is supplied to a block lacking detecting circuit 15, k-th lacking is detected, and by an output ER, a switching signal generating circuit 16 switches a switch 13 by a signal SW and inserts a block obtained by an interpolation processing as substitute for a defective block. By the signal ER, an extracting circuit 20 inputs k - 1th and k + 1-th blocks to turn-back circuits 21, 22 and inverts a time series of an input data block, and thereafter, multiplies weighting coefficients k1, k2 by weighting circuits 23, 24, and executes addition by an adding circuit 27. An obtained interpolation block is data varied gradually from the k - 1 block to the k + 1 block, therefore, it has a roughly average spectrum, no discontinuity is generated in a joint and a satisfactory reproducing sound is obtained.

Description

[Detailed description of the invention] [Industrial application field]

The present invention relates to an audio signal interpolation device.

[Summary of the invention]

The present invention extracts the audio signals of the blocks before and after the defective block when the input audio signal is divided into blocks by predetermined time lengths, and reverses the time series of the audio signals of the blocks before and after the defective block. After applying a predetermined weight to each of them, they are added together to form an interpolation block, and this interpolation block is taken out as an output signal in place of the defective block. No noise is generated at the joints, and good reproduction sound and voices can be obtained.

[Conventional technology]

For example, in digital audio signal playback devices such as CDs (compact discs) and DATs (digital audio tapes), errors may occur in the playback signals due to various factors such as disc scratches, scratches on the tape, and contact between the tape and the head. occurs. Therefore, the digital audio signal is normally recorded after being subjected to error correction encoding processing, and the errors are corrected by performing error correction decoding processing on the playback device. However, error correction codes have a limited correction ability, and when an error exceeding the correction ability occurs, data is corrected by interpolation processing. This interpolation process utilizes the correlation that the audio data itself has, as shown in Figure 5 A, B, and C.
Conventionally, pre-hold, average value interpolation, linear interpolation, etc. have been used. In FIG. 5, the correct sample data is indicated by 0, and the sample data indicated by 0 is the sample data generated by interpolation processing. As is clear from the figure, pre-hold uses immediately preceding data instead of error data, and average value interpolation obtains the average value of correct sample data before and after error data instead of error data. In addition, linear interpolation is used when there are multiple consecutive error samples, for example when there is a signal dropout such as a dropout, and the data in the error interval is calculated from the sample data before this error interval to the sample after the error interval. It is reproduced assuming that the data changes linearly.

[Problem to be solved by the invention]

However, when errors occur in a relatively large amount of consecutive data, resulting in a relatively long defect, the above-mentioned conventional interpolation process may produce abnormal sounds at the defective portion, making it undesirable for the auditory sense. do not have. For example, when compressed audio data is handled, such as in the CD-1 format or magnetic disks, relatively long defects as described above often occur. That is, for example, in the CD-1 format, data is recorded on a disc as a plurality of sectors, but when the data is audio, the sector structure is as shown in FIG. In other words, one sector consists of 12
A sound group 5G-00~ consisting of a synchronization signal 5YNC for bytes, a header section for 4 bytes, a subheader section for 8 bytes, and 18 pieces, each consisting of 128 bytes.
It consists of 5G-17, a 20-byte space, and a 4-byte reserve section. So. One sound group in the case of level B stereo is structured as shown in FIG. That is, in the same figure, SP' (j=0 to 7) is the sound parameter, Dm, + (k-0 to 27) is the sound data,
SU' indicates a sound unit. The sound data Dh,4 is 4-bit compressed data (compression ratio 1/4), and one sound unit is composed of 28 pieces of sound data in the vertical direction in the figure. Note that j is a sound unit number and k is a data sampling number. In the case of this CD-1 format, if an error occurs in the sound parameters, the data for one sound unit or the entire sound group becomes unusable. The data for the time of 2.97m5ec cannot be reproduced.Also, if the synchronization signal 5YNC cannot be detected or the header or subheader cannot be extracted, the data of the entire sector becomes unusable, and in this case, the data of 53. Data becomes unreproducible over a period of 5m5ec. If a defect occurs where data cannot be reproduced over a long period of time, the correlation between data before and after the defect period will be low, so the pre-hold or linear In the conventional interpolation method for interpolation lights, abnormal sounds occur in the defective parts. Therefore, for example, as shown in FIG. 6A, a data block consisting of a plurality of samples is )
..., and when the block cannot be reproduced, a method can be considered to repeat the previous (k-1) blocks, but in this case, the joints between the blocks are generally not continuous and there are no differences. It generates sound. Also, Figure 6B
As shown in , it is also possible to discard a block that cannot be reproduced due to a defect and perform a so-called cross-fade between the previous and subsequent (k-1) blocks and (k+1).
This method has the disadvantage that the time axis is shortened. In view of the above points, the present invention aims to provide an interpolation device that does not generate abnormal noise and does not involve changes in the time axis.

[Means to solve the problem]

This Φ invention includes a defect detection means for detecting a defective block when an input audio signal is divided into blocks by predetermined time lengths, and a defect detection means for detecting a defective block based on the detection output of the defect detection means. Preceding and subsequent block extracting means for extracting the preceding and succeeding blocks, and -1 weighting for the signals obtained by reversing the time series of the audio signals of the extracted preceding and succeeding blocks, respectively;
The apparatus includes interpolation processing means for adding both, and means for converting the output of the interpolation processing means into an output signal in place of the audio signal of the block having the defect.

[Effect]

In this invention, when a defective block is detected, the defective block is reproduced from the audio signals of the blocks before and after the defective block. At that time, the data of each block before and after is reversed in time series, weighted and added together to reproduce the defective block.

【Example】

Hereinafter, an embodiment of an audio signal interpolation device according to the present invention will be described with reference to the drawings. This example is an example in which the present invention is applied to level B stereo PCM audio data in the CD-1 format described above, and in this example, one sound unit (28 samples) is treated as one data block. In addition, 1
Of course, a sound group may be treated as one data group, or one sector may be treated as one data block. FIG. 1 is a block diagram of the interpolation device of this example, which can actually be processed by software using a microcomputer. FIG. 2 is a time chart for explaining this example. Now, as shown in FIG. 2A, the data block is (k
-2)th, (k-1)th, kth, (k-1)th, etc. It is assumed that PCM data DAo is supplied to the input terminal 11. This human-powered PCM data DAo
is supplied to the two-block delay circuit 12 through the input terminal 11 and delayed by two blocks to become two-block delayed PCM data DA2 (H in FIG. 2), which is supplied to one input terminal of the switch circuit 13. be done. The input PCM data DAo is also supplied to the one block delay circuit 14 through the input terminal 11 and delayed by one block, resulting in one block delayed PCM data DAI (FIG. 2B). This one block delayed PCM data D
A is supplied to the block missing detection circuit 15. Here, if the second data block of the input PCM data DA is missing, for example, because the sound parameter SPl cannot be reproduced, as described above, the block missing detection circuit 15 detects the second data block. It is detected that a data block is missing, and a detection output ER (C in the same figure) is obtained. This detection output ER is supplied to the switching signal generating circuit 16, and as described later, the switching signal generating circuit 16 outputs the interpolated data block obtained by the interpolation process in this example, and the second data block that is missing. Instead, a switching signal SW (D in the figure) for switching and inserting into the PCM data DA2 is obtained. This switching signal SW is
The signal is supplied to the switch circuit 13 as a switching control signal. The input PCM data DAo is further supplied to the preceding and following block extraction circuit 20 through the input terminal 11, and
The detection output ER from the block missing detection circuit 15 is supplied to the preceding and following block extraction circuit 20. This preceding and following block extracting circuit 20 is equipped with a memory having a capacity to store three data blocks, and the detection output ER is
When supplied to 0, the (k-1)th data block before the missing second data block (the one in the middle of the three stored data blocks) and the next data block are (k+1)th data block is obtained from this. (
The k-1)th and (k+1)th data blocks are
They are supplied to folding circuits 21 and 22, respectively. These folding circuits 21 and 22 perform a process of reversing the time series of the input data blocks. That is, the data strings of the (k-1)th and (k+1)th blocks are expressed as
+ s (n -0, 1, 2-(N-1); N is the number of samples in 110-), the folded data blocks R (k-1) and R that are the output data
(k+1), data string X *-H, N-++-1 and X &+1. □
-1 is obtained (see Figure 3). By the way, as a basic property of Fourier transform, J[f (
-t>] -f-f (-t) e -'“'dt=
-Im-f (x) e -''''dx-Lctr f
(x) e −”−”” dx−F(−ω) where,;' [f (t) ] −F (ω
) = f-oo t (t) e'-''''di (J indicates Fourier transform). Also, when f (t) is a real number, the absolute value spectrum IF
(ω)1 is an even function of ω, and the phase spectrum is an odd function. Because F(-ω)-F (ω) Therefore, F (-ω) l e"'-"' -l F ((I
J) l e-””,', l F (-ω)l-I
This is because F(ω)Φ(−ω)■−Φ(ω). Therefore, the folded data string X &-1. □, -1 is (
It has the same absolute value spectrum as the data sequence X, -, , of the k-1)th block. Also, the wrapped data string
k+1. N-11-1 has the same absolute value spectrum as the data string Xhbr,s of the (k+1)th block. The folded data block R(k-1) and R(k+l) columns from the folding circuits 21 and 22 are supplied to weighting circuits 23 and 24, respectively, which are formed by multiplication circuits. 2 is supplied to the weighting circuits 23 and 24, the weighting coefficients from the weighting coefficient generating circuits 25 and 26, and 2 to the folded data string, as described later.
Weighting is applied. In this case, the weighting for the (k-1)th data block is such that + is added to the coefficient, and the coefficient data string Wk is weighted. When expressed as , the first value Wk, o is "1", the subsequent values are "1" or less, and the final value W*, s-+ (W5.2 in this example) is "0''. The coefficient 2 is set to the coefficient data string W*, N-n-1 in the reverse order. In this example, the weighting may be a coefficient, and the variable weighting may be a function as shown in FIGS. 4A, B, and C, for example. For example, when weighting is applied to the coefficients, K2 having a linear change as shown in A in the figure is used, w, -1a-n where a-1/(N-1). The output data of these weighting circuits 23 and 24 are then supplied to an adder circuit 27, where they are added together to obtain an interpolated data block k (interpolated) (FIG. 2G). This interpolated data block k (interpolation) gradually reduces the folded data R(k-1) of the (k-1)th data, while the folded data R of the (k+1)th data
(k+1) is gradually increased and the two are added together. In this case, as mentioned above, the weighting is by the coefficient K. The leading value of is "1", and the final value of the (k-1)th data block and the leading value of its folded data block R(k-1) are the same, so (k-1
)-th data block and interpolation data block k (interpolation) become smooth. Also, the final value of the weighting coefficient of 2 is "1", and the starting value of the (k+1)th data block and the final value of its folded data block R(k+1) are the same, so (k
+1)th data block and interpolated data block k
(Interpolation) will also become smoother. In other words, interpolated data block k (interpolated) is a cross-fade of a data sequence that has the same absolute value spectrum as the data block before the missing data block and a data sequence that has the same absolute value spectrum as the subsequent data block. As a result, it has a spectrum that is almost the average of the blocks before and after it, and the connections between the blocks are smooth and connected without causing any discontinuity. The interpolated data block k (interpolated) thus obtained is supplied to the other input terminal of the switch circuit 13. The switch circuit 13 receives a switching signal SW from the switching signal generation circuit 16.
Therefore, the two-block delay signal DAo from the delay circuit 12
In the second data block period (FIG. 2H), the interpolation data k (interpolation) from the adder circuit 27 is switched. Therefore, as shown in FIG. 2, the output terminal 17 receives interpolated data k instead of the k-th data block.
PCM data DA with (interpolation) inserted is obtained. As mentioned above, this output PCM data DA has smooth joints between the interpolated data block and the blocks before and after it, so no harsh noise is generated at the joints during audio playback. Furthermore, since missing data blocks are replaced with interpolated data blocks, the time axis direction is not shortened. Note that the interpolated data k(
The process of obtaining interpolation) from the data blocks before and after it is as follows:
Practically, the arithmetic processing shown by the following equation can be performed by microcomputer processing. X & n -W k, n X X *-1, N-1
1-1+ W 叡N-n-r X X k+1. N-
e-1 This operation is not a complicated operation like an FFT operation, but simply swaps the preceding and following block series, and weights are multiplied by a coefficient and added, so expensive hardware is not required. It has the advantage of not requiring calculations such as irrational numbers. Note that the above example takes playback of PCM audio data in CD-1 format, so we considered sound units, sound groups, sectors, etc. as processing blocks for audio signals. It goes without saying that even if the sample does not have a structure, the present invention can be applied by treating a collection of several samples as a block for convenience. Further, the weighting coefficient change function is not limited to that shown in FIG. 4, and various functions can be used. In addition, the block missing detection means detects not only errors in all of the data in one block, but also in cases where a small number of errors that cannot be corrected by the error correction code occur in the data in one block. It may be detected as data missing. Further, although the above example describes interpolation processing of PCM audio signals, the present invention can also be applied to analog audio signals.

【Effect of the invention】

As described above, according to the present invention, since interpolated data is formed from the data strings of the blocks before and after the missing block, no change in the time axis direction occurs in the output data after the interpolation process. Then, the interpolated data uses a data sequence that is a reversed time series of the data before and after the missing block, and gradually moves from the data of the block before the missing block to the data of the block after the missing block. Since the data changes, it has a spectrum that is almost the average of the data blocks before and after it, and there is no discontinuity at the joint between the interpolated data and the block data before and after it, and good playback sound is obtained. That's what you get.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of an embodiment of an interpolation device according to the present invention, FIG. 2 is a flowchart for explaining its operation, FIG. 3 is a diagram for explaining its operation, and FIG. 4 is a weight 5 is a diagram for explaining a conventional interpolation method, FIG. 6 is a diagram for explaining another conventional interpolation method, and FIG. 7 is a diagram for explaining a conventional interpolation method. FIG. 8 is a diagram showing the sector structure for I format audio data.
It is a diagram showing the structure of the sound group. 11; Audio data input terminal 12.14; Delay circuit 13; Switch circuit 15; Block missing detection circuit 16; Switching signal generation circuit 17; Audio data output terminal 20; Block extraction circuit 21.22 before and after the missing block
; Folding circuits 23 and 24; Weighting circuit 27; Addition circuit

Claims (1)

[Claims] Defect detection means for detecting a defective block when an input audio signal is divided into blocks by predetermined time lengths; preceding and following block extracting means for extracting blocks before and after; interpolation processing means for weighting signals obtained by reversing the time series of the audio signals of the extracted preceding and following blocks, respectively, and then adding both; 1. An audio signal interpolation device comprising: means for replacing the output of the processing means with the audio signal of the defective block as an output signal.