JPH0383421A - Signal forecasting encoder and decoder - Google Patents

Abstract

PURPOSE:To simplify a manufacture process and to improve the rate of satisfactory products by repating operation to write quantized data into a memory, afterwors, to read out these data from the memory and to respectively generate the forecasting signal of the next sampling point by signal forecast encoding and decoding devices based on the data. CONSTITUTION:A quantizing output A of a signal forecast encoding device 11 is written into a memory 12 and afterwards, these written data or previously written data A' are read out. At such a time, when there is a defect in the memory 12, the above mentioned data A' are not made same as the data A quantized by the signal forecast encoding device 11. In such a case, since a forecast value at the next sampling point is generated in the signal forecast encoding device 11 so that a differential signal can be made minimum, the forecast value is encoded to cancel deviation. Namely, the bit error of the data A' read out from the memory 12 is corrected by a feedback loop as forecasting in the signal forecast encoding device 11, the deviation in the output signal of a decoder 13 is canceled after the lapse of a certain time.

Description

[Detailed Description of the Invention] [Object of the Invention] (Industrial Application Field) The present invention relates to a signal encoding/decoding device that encodes an audio signal, stores it in a digital memory, and then decodes it.
In particular, the present invention relates to a signal predictive coding/decoding device using a signal coding device using a predictive coding method.

(Prior Art) FIG. 8 shows a conventional signal predictive coding/decoding device, in which 81 is a signal predictive coding device, 82 is a memory, and 83 is a signal predictive coding device.
is a decoding device, and the signal code 11!1J encoding device 81
The quantized output A is stored in a memory 82, and this stored data is read out and restored by a decoding device 83 to a signal close to the original input signal.

When encoding an analog input signal into digital data, the signal predictive encoding device 81 uses ADM (adaptive delta modulation), ADPCM (adaptive differential pulse code modulation), or APC.
(Adaptive predictive coding) is used to perform coding based on the state fi (or characteristics thereof) of the human input signal and past input signals.

Figures 9 to 11 are, for example, the first volume of the Electronic Information and Communication Handbook (edited by the Institute of Electronics, Information and Communication Engineers; published by Ohmsha),
Excerpts are shown of various signal predictive coding devices shown in the category "Coding of speech/acoustic signals", and Fig. 9 shows A
DM encoding system signal encoding device, Figure 10 is ADPC
FIG. 11 shows a signal encoding device using the M encoding method. FIG. 11 shows a signal encoding device using the APC method. In addition to these, S.B.
Many predictive coding methods have been introduced, such as a C (subband coding) method signal coding device.

Furthermore, the Institute of Electronics, Information and Communication Engineers, April 1987 issue 4-2,
Various predictive coding methods are also introduced in ``Medium/High Bitrate Speech Coding''.

In addition, in FIG. 9, X is an input signal, 91 is a subtracter,
92 is a 1-bit quantizer, A is the quantized output of the 1-bit quantizer 92, 93 is a Δ (quantization level) variable device, 94 is a multiplier, 95 is an integrator, y is a prediction signal, x-y is a differential signal.

In FIG. 10, X is an input signal, 101 is a subtracter, 1
02 is an adaptive quantizer, A is the quantized output of the adaptive quantizer 102, 103 is an adaptive inverse quantizer, 104 is an adder, 10
5 is an adaptive predictor, y is a prediction signal, and xy is a difference signal.

In FIG. 11, X is an input signal, 111 is a subtracter, 1
12 is an adaptive quantizer, 113 is an inverse quantizer, 114 is an adder, 115 is an adaptive predictor, y is a prediction signal, x-y is a difference signal, 116 is a linear prediction analyzer, 117 is a multiplexer, A is a This is the quantized output selected by multiplexer 117.

The operations of the signal predictive coding devices shown in FIGS. 9 to 11 are described in the above-mentioned documents and their references, so a detailed explanation will be omitted here, but using the quantized output A, the following The feature is that it generates a preliminary JIIJ value y at the time point (sample point) (this operation also applies when the noise shaper is inserted).

By the way, the conventional signal element δ1 encoding/decoding device shown in FIG. However, if there is a defect in the memory 82, the data A read from the memory 82 will no longer be the same as the quantized output A of the signal coder 8111 encoder 81, and the prediction in the signal predictive encoder 81 will be decoded. A difference will occur between the signal y and the output signal y of the decoding device 83.

In this case, the signal predictive encoding device 81 generates a code for restoring the next time point (sample point) without knowing the occurrence of the above-mentioned difference. Therefore, the decoding device 8
There is a risk that the output signal (predicted signal) y of No. 3 will deviate greatly from the original input signal X.

Here, as the simplest example, A
The operation of a signal predictive encoding/decoding device using a signal encoding device using the DM encoding method will be briefly described. now,
If the quantization output A is “1”, the predicted value at the previous point)' a
It is defined that the amplitude is made larger by the difference value Δ than -+, and conversely, if the quantized output A is "O", the amplitude is made smaller by the difference value Δ than the predicted value y, -1 at the previous point in time. . At this time, the DC level of the quantized output A is ro 1010101
...'', and this code string is stored in the memory 82. If memory 82 is defective, memory 82
Suppose that there is a 1-bit error in the code string of data A read from the decoder 83 and used by the decoding device 83, for example, "rollolol...".

Here, the predicted signal y generated from the original code string by the signal predictive coding device 81 is shown in FIG. The resulting output signal y is shown in FIG. 12(b). From this figure, it can be seen that when an error occurs, the DC level of the output signal y of the decoding device 83 deviates, and it does not return to the DC level before the error even after that. Due to this, there is a possibility that the output signal y of the decoding device 83 may jump out of the dynamic range or that transient noise may become large.

As described above, if there is an error in the data used by the decoding device 83 due to a defect in the memory 82, there is a risk that the output signal y of the decoding device 83 will be seriously impaired. Furthermore, in general, the larger the capacity of a semiconductor memory, the more microfabrication techniques are used and the manufacturing process becomes more complex, so the yield rate decreases. Techniques such as redundant circuits are being used to increase the rate of non-defective memory, but in the end, this is one of the major causes of increased costs. .

(Problem to be solved by the invention) As mentioned above, the conventional signal predictive encoding/decoding device
If there is a defect in the memory for storing quantized data by the signal predictive coding device, the data read from the memory will not be the same as the quantized data, and the predicted signal in the signal predictive coding device and the decoding device will be damaged. There is a problem in that the output signal of the decoding device deviates greatly from the original input signal, causing a serious problem.

The present invention has been made to solve the above-mentioned problems, and the purpose is to solve the problem when the memory for storing quantized data by the signal predictive coding device is defective, and the data read from the memory is not quantized. To provide a signal predictive encoding/decoding device in which an output signal of a decoding device does not deviate greatly from an original human-powered signal and does not cause serious trouble even if bits are no longer the same as data.

[Structure of the Invention] (Means for Solving the Problems) The present invention writes data obtained by quantizing an analog input signal by a signal predictive coding device into a memory, and data read from this memory is quantized by a decoding device. In a signal predictive encoding/decoding device that restores a signal close to an input signal, after writing the quantized data to a memory, reading this written data or previously written data from the memory, The present invention is characterized in that the signal predictive encoding device and the decoding device each repeat the operation of generating a predicted signal for the next sample point based on the read data.

(Operation) If there is no defect in the memory, the data read from the memory will be the same as the data quantized by the signal predictive coding device, and the signal predictive coding device and the decoding device will perform the signal predictive coding device and the decoding process based on this same data. Since each device generates a predicted signal, it becomes possible for the decoding device to restore a signal close to the original input signal, as in the past. On the other hand, if the memory is defective, the data read from the memory will not be the same as the data quantized by the signal precoder 111 encoder. At this time, when the decoding device generates a predicted signal based on the data read from this memory, it temporarily outputs a signal that deviates from the restored signal that should be output. Furthermore, when the signal predictive coding device generates a predictive signal based on the data read from the memory, it outputs a signal that deviates from the predictive signal that should originally be output. in this case,
In a signal predictive coding device, a predicted value at the next time point (sample point) is generated in order to minimize the difference signal (x - y), so it is encoded to cancel out the deviation, and then decoded after a certain amount of time has elapsed. This will eliminate the deviation in the output signal of the device.

(Example) Hereinafter, an example of the present invention will be described in detail with reference to the drawings.

FIG. 1 shows a signal predictive encoding/decoding device,
11 is a signal predictive encoding device, 12 is a memory, and 13 is a decoding device. Data is read out, and based on the read data A, the signal predictive encoding device 11 and the decoding device 13 each generate predictive signals. The signal predictive coding device 11 uses algorithms such as ADM (adaptive delta modulation), ADPCM (adaptive differential pulse code table, S1), and APC (adaptive predictive coding) when coding an analog input signal into digital data. The state of the input signal and past human input signal! ! ! (or its characteristics).

FIG. 2 shows A as the signal predictive coding device 11 in FIG.
This shows an example of a case where a signal encoding device using the DM encoding method is used, and compared to the conventional signal encoding device using the A'DM encoding method shown in FIG.
The quantized output A of is written to the memory 12, and the data A obtained by reading this written data or previously written data (for example, 3 sample points) from the memory 12 is input to the decoding device 13. The highlight is that instead of the quantized output A, the operation of feeding back to the Δ variable device 93 and multiplier 94 to generate the predicted signal y of the next sample point is repeated.

FIG. 3 shows A as the signal predictive coding device 11 in FIG.
This shows an example of a case where a signal encoding device using the DPCM encoding method is used, and compared to the signal encoding device using the conventional ADPCM encoding method shown in FIG.
The quantized output A of 2 is written to the memory 12, and the data A obtained by reading this written data or previously written data (for example, 3 sample points) from the memory 12 is input to the decoding device 13. The difference is that instead of the quantized output A, the operation of feeding back to the adaptive inverse quantizer 103 to generate the predicted signal y of the next sample point is repeated.

FIG. 4 shows A as the signal predictive encoding device 11 in FIG.
This shows an example in which a PC-based signal encoding device is used, and compared to the conventional APC-based signal encoding device shown in FIG. Data A obtained by reading this written data or previously written data (for example, 3vA main point) from the memory 12 is input to the decoding device 13, and is demultiplexed by the demultiplexer 41. , this demultiplexed signal is fed back to the inverse quantizer 113 in place of the quantized output A, and fed back to the adaptive predictor 115 in place of the prediction coefficient output of the linear prediction analyzer 116 to generate the next The difference is that the operation of generating the predicted signal y of the sample point is repeated.

In the signal predictive encoding/decoding device of each of the above embodiments, if the memory 12 is free of defects, the data A read from the memory 12 will be the same as the data A quantized by the signal predictive encoding device 11, Since the signal predictive coding device 11 and the decoding device 13 each generate a predicted signal based on this same data, it becomes possible for the decoding device 13 to restore a signal close to the original human signal. On the other hand, if the memory 12 is defective, the data A read from the memory 12 will not be the same as the data A quantized by the signal predictive encoding device 11. At this time, when the decoding device 13 generates the pre-1111J signal based on the data A read from the memory 12, it temporarily outputs a signal that deviates from the restored signal that should originally be output. Furthermore, when the signal predictive encoding device 11 generates a predictive signal based on the data A read from the memory 12, it outputs a signal that deviates from the predictive signal that should originally be output.

In this case, the signal predictive encoding device 11 uses the difference signal (x
Since a predicted value at the next time point (sample point) is generated in order to minimize the difference (y), encoding is performed so as to cancel out the deviation. That is, 1. Bit errors in data A read from memory 12 are corrected through a feedback loop of prediction in signal predictive encoding device 11, and the deviation in the output signal of decoding device 13 is eliminated after a certain period of time has elapsed. Therefore, even if there is a defect in the memory 12, an operation is performed to cancel the effect caused by the defect. Here, the solid line in FIG. 7 shows an example of the movement of the DC level of the output signal of the decoding device 13 when the bit error occurs, and for comparison, the conventional signal predictive coding The movement of the DC level of the output signal of the decoding device in the decoding device is shown by a dotted line in FIG.

FIG. 5 shows A as the signal predictive coding device 11 in FIG.
Another example is shown in which a PC type signal encoding device is used, and the quantization output (bit division) of the adaptive quantizer 112 is different from the conventional APC type signal encoding device shown in FIG. quantized output) Al and multiplexer 1
Quantization output A2 selected by 17 and weak to bit errors
and are processed differently. That is, the quantized output A1, which is resistant to bit errors, is written to the first memory 51 where there may be a defect such that the data written at the same address and the data read are different, and the written data or the previous data are Written data (for example, 3 sample points)
The data A1 obtained by reading out from the memory 51 is input to the decoding device 13, and is fed back to the inverse quantizer 113 instead of the quantized output A1 to generate the predicted signal y of the next sample point. I have to. On the contrary,
The quantized output A2, which is susceptible to bit errors, is written to the second memory 52 that is known to be defect-free, and this written data or previously written data (for example, 3 sample points) is stored in the memory. The data A2 obtained by reading from the decoding bag rI113 is inputted to the decoding bag rI113. The decoding device 13 includes two memories 51 as described above.
．． 5. Predicted signal y of the next sample point from the readout output from 2
generate.

Note that the above-mentioned quantized output A2 that is susceptible to bit errors refers to a case where the predicted signal y deviates from the original value to an unacceptable degree if the data read from the memory 52 is no longer the same as the quantized output A2. refers to the sign that occurs.

Furthermore, although the two memories 51 and 52 are shown separately in FIG. 5, this means that the address space of one memory may be divided into two.

FIG. 6 shows a general configuration of a signal predictive coding decoding device using two memories in the signal predictive coding device shown in FIG. 5, and 11 is a signal predictive coding device. 61 is a first memory known to be defect-free, 62 is a second memory, and 13 is a decoding device. Here, after writing data to the second memory 62 only for the quantized output A2 that is susceptible to bit errors, this written data or previously written data is read from the second memory 62, Based on this read data A2, the signal predictive coding device 11 and the decoding device 13 repeat the operation of generating a predicted signal for the next sample point, and the signal coding device 11 → the first memory 61
The data flow along the path of the -4 decoding device 13 is the same as in the conventional case.

Furthermore, even when using an APC signal encoding device as shown in FIG. 4, the memory area is divided into two, and a code that is vulnerable to memory defects and a code that is resistant to memory defects are stored separately. Good too.

[Effects of the Invention] As described above, according to the present invention, if the memory for storing quantized data by the signal predictive coding device has a defect, the data read from the memory will not be the same as the quantized data. Even if this occurs, it is possible to realize a signal predictive encoding/decoding device that performs an operation to cancel the effect caused by this, and in which the output signal of the decoding device does not deviate significantly from the original input signal.

Therefore, the present invention can be applied to an answering machine that digitally records voice messages using, for example, DRAM (dynamic random access memory), and that can be used to convey the message to the other party when a call comes in, or to record the message of the other party. If applied to the encoding/decoding of audio equipment in telephones, etc., even if the memory has bit defects, the effects of defects (for example, the effect on sound quality) can be suppressed to a small extent, thereby increasing the quality of memory. It is no longer necessary to use redundant circuits to increase the memory capacity, making it possible to reduce the cost of memory and the cost of the device itself.

[Brief explanation of drawings]

FIG. 1 is a configuration explanatory diagram showing an embodiment of a signal predictive coding/decoding device of the present invention, and FIG. 2 is a configuration explanatory diagram showing an embodiment of a signal predictive coding/decoding device of the present invention. FIG. 3 is an explanatory diagram showing an example of a configuration in which a signal encoding device of the ADPCM encoding method is used as the signal predictive encoding device in FIG. 1, and FIG. A configuration explanatory diagram showing an example of a case where an APC type signal encoding device is used as the signal predictive encoding device in Fig. 1, and Fig. 5 is an APC type signal encoding device as the signal predictive encoding device in Fig. 1. FIG. 6 is a diagram showing a general configuration of a signal predictive encoding/decoding device using two memories in the signal predictive encoding device, and FIG. Figure 8 shows an example of the movement of the DC level of the output signal of the decoding device when a bit error occurs in the memory in the signal predictive encoding/decoding device of the present invention. Configuration explanatory diagram showing the converting device, No. 9
The figure is a configuration explanatory diagram showing a signal 7:) differentiation device of the conventional ADM encoding method, FIG. 10 is a configuration explanatory diagram showing a signal encoding device of the conventional ADPCM encoding method, and FIG. FIG. 12(a) is a diagram showing the predicted signal y generated by the signal predictive encoding device in FIG. 8, and FIG. FIG. 3 is a diagram showing an output signal y generated by the decoding device when the memory shown in the figure is defective and there is a 1-bit error in the code string of data used by the decoding device. 11... Signal predictive coding device, 12,51°62...
・Memory that may be defective, 13...Decoding device, 41...Demultiplexer, 52°61...Memory known to be non-defective, 91...Subtractor, 92 ... 1-bit quantizer, 93 ... Δ (quantization level) variable device, 94 ... multiplier, 95 ... integrator, 101 ... subtractor, 102 ... adaptive quantization vessel, 1
03...Adaptive inverse quantizer, 104...Adder, 10
5...Adaptive predictor, 111...Subtractor, 112...
・Adaptive quantizer, 113... Inverse quantizer, 114...
Adder, 115...Adaptive predictor, 116...Linear prediction analyzer, 117...Multiplexer.

Claims (3)

[Claims] (1) Signal predictive coding/decoding in which the analog input signal is quantized by a signal predictive coding device, the data is written into memory, and the data read from this memory is restored to a signal close to the original input signal by a decoding device. In the encoding device, after writing the quantized data into the memory, this written data or previously written data is read out from the memory, and the signal predictive encoding device and a signal predictive encoding/decoding device, wherein the decoding device repeats the operation of generating a predicted signal for each next sample point. (2) For memory that may have a defect where the data written at the same address and the data read are different, a quantized output that is resistant to bit errors is written, and it is determined that there is no defect. 2. The signal predictive encoding/decoding apparatus according to claim 1, wherein the quantized output, which is susceptible to bit errors, is written into the memory. (3) Only for quantized outputs that are susceptible to bit errors, after writing data to the memory, this written data or previously written data is read from the memory, and based on this read data, Then, the signal predictive coding device and the decoding device each generate a predicted signal for the next sample point, and based on this data, the signal predictive coding device generates the next predicted signal for the quantized output that is resistant to bit errors. generate a predicted signal for the sample point, and after writing the data to a memory known to be defect-free, read this written data or previously written data from the defect-free memory. The signal predictive encoding/decoding device according to claim 1, wherein the decoding device generates a predicted signal for the next sample point based on the read data.