JPH04171499A - Adaptive differential pcm decoder - Google Patents

Abstract

PURPOSE:To obviate large variation in amplitude absent in the original sound wave-form of decoded waveform by outputting a control signal associated with the generation of a carry signal or borrow signal to judge controllably whether or not a calculating output of an adder subtracter should be taken in by a register. CONSTITUTION:While amplitude data 103 and code data 104 are inputted to an adder-subtracter 6 to be outputted as calculating output 105, a gate control signal 106 formed by inverting a carry signal or borrow signal is outputted. The calculating output 105 is inputted to an X register 7, while a gate control signal 106 is sent to an AND circuit 10. A latch signal 107 for the X register 7 is also inputted to the AND circuit 10 and to the X register 7 only in the high level of the gate control signal 106 and DA converted in a DA converter 8 to be outputted as an analog signal 108. Thus, the generation of variation in amplitude absent in the original sound waveform of decoded wave form of ADPCM(Adaptive differential PCM) code is obviated.

Description

Detailed Description of the Invention [Field of Industrial Application] The present invention relates to an ADPCM decoder, and particularly to an ADPCM decoder for decoding ADPCM codes used in the field of speech synthesis.
Regarding a CM decoder.

[Conventional technology]

ADPC conventionally used in the field of speech synthesis
As shown in FIG. 4, the M decoder includes a register 1 that latches the ADPCM code data 108, and a decoder 2 that decodes the value of the L register 1 to determine the amount of movement of the quantization width pointer 110.

The adder/subtracter 9 and the output of the adder/subtracter 9 are set to the quantization width pointer 1.
The limiter 3 suppresses the value within the upper and lower limits of 10, the A register 4 latches the value of the quantization width pointer 110, the address is selected by the L register 1 and the A register 4, and the difference of the next sampling point is selected. The difference value data table ROM 5 outputs the value, and the value of the X register 7 latches the result obtained by sequentially accumulating the difference values.
An adder/subtractor 6 which adds or subtracts the output of the differential value data cheaply ROM 5 to calculate the amplitude value of the next sampling point, and an analog signal 115 which converts the output of the X register 7 into an analog signal. DA converter 8 to output
It is composed of:

ADPCM (Adaptive different)
The 4) method is a data compression method that utilizes the strength of the correlation between adjacent samplings of audio, and is a method in which the difference in amplitude values at consecutive sampling points is expressed as a compression code.

In the following, the ADPCM code is a 4-bit code consisting of 1 bit representing positive/negative polarity and 3 bits representing amplitude, and the quantization width pointer is assumed to be 3 bits, and the current quantization width pointer and ADPCM code are assumed to be 3 bits. The difference value data selected by the amplitude data of is added or subtracted from the amplitude value of the current sampling point according to the value of the polarity bit of the ADPCM code to calculate the amplitude value of the next sampling point, and the current quantization A configuration will be described in which instead of performing a multiplication process in which a value obtained by multiplying the width by a prediction coefficient is used as the quantization width of the next sampling point, the multiplication results are stored in advance in a ROM as a table as difference value data.

First, before decoding the ADPCM code, the initial value 0 is input to the A register 4, and the midpoint data 80H of the output waveform is input to the 8-bit X register 7, which accumulates amplitude value data. I'll keep it. To determine the amplitude value at the first sampling point, register L, register 1f17) A
By inputting the DPCM code data 109 and the value of the quantization width pointer 110 of the A register 4 as the address of the difference value data table ROM5, the amplitude data of the ADPCM code and the current quantum Amplitude data 111 including amplitude data of a maximum of 8 bits of difference value determined by the conversion width, and 1-bit positive/negative sign data 112 are obtained.

The contents of the amplitude data (HEX data) in the difference value data table ROM 5 corresponding to the long values of the L register 1 and the A register 4 are as shown in Table 1 below. Regarding the contents of the value data, it is assumed that adding or subtracting the value of the A register 4 is equivalent to multiplying the difference value data by a prediction coefficient.

The amplitude data 111 of the output of the X register 7 and the difference value data table ROM 5 is determined by the value of the positive/negative sign data 112.
are added or subtracted in the adder/subtractor 6, and the resulting calculation output 113 is latched in the X register 7, and D
The A converter 8 converts it into an analog signal and outputs it.

After setting the latch signal 114 in the X register 7 to low level and latching the value of the The output of 2 and the value of the A register 4 are added or subtracted in the adder/subtractor 9, and if the result of the operation is negative in the limiter 3, 0 is input to the A register 4, and if the result of the operation is a value of 8 or more, , 7 are input to the A register 4.

[Table 2] In this way, values from 0 to 7 are input to the A register 4, and are used as the quantization width pointer 110 to select the difference value data multiplied by the prediction coefficient of the next sampling point.

As explained above, each time new ADPCM code data is input, the difference value is selected between the value of L register 1 and the value of A register 4 as the address of the difference value data table ROM5, and the value is stored in the X register 7. The amplitude value of the next sampling point is obtained by adding or subtracting from By repeating the operation of adding or subtracting the value decoded by the decoder 2 to the A register 4 and updating the value of the quantization width pointer 110, the decoding process of the ADPCM code data 1 data is performed.

[Problem to be solved by the invention]

In the above-mentioned conventional ADPCM decoder, as shown in the waveform diagram of original sound encoded data in FIG. After calculating the sampling value S2 by adding Δ2, when calculating the next sampling point, if the value of the quantization width pointer is large, the absolute value of the difference value data will change greatly, and the ADPCM There are cases where only Δ, ゛ and Δ3 can be selected as the amplitude data of the sign, and the error from the original sound is Δ, °
Choosing Δ3 results in a smaller value than Δ3.

In order to reduce the error with the original sound waveform, select Δ and
Even if DPCM encoding is performed, since the ADPCM code is data obtained by compressing the difference values, even if the amplitude data obtained by accumulating the difference values exceeds the maximum value FFH, it is possible to encode the data as is.

On the other hand, if only Δ9° and Δ9 can be selected as the amplitude data of the ADPCM code at the sampling point, the error with the original sound will be smaller if Δ9 is selected than Δ9'. In order to reduce the error from the original sound waveform, select Δ9 and select A.
Even with DPCM encoding, even if the amplitude data obtained by accumulating the difference values becomes smaller than the minimum value O, it is possible to encode it as is.

In this way, when the ADPCM code data obtained by encoding the amplitude data from Sl to StZ in FIG. 5 is reproduced by the conventional decoder in FIG. 4, as shown in FIG. The added value becomes X2, and Δ is added to X2.
, the added value exceeds FFH, so ×
3 is a value close to O. - On the other hand, the value obtained by subtracting Δ8 from X7 becomes X8, and when Δ9 is subtracted from x8, a borrow occurs, and the subtracted value X9 becomes a value close to FFH. There is a drawback that the amplitude value of x9 is completely different from the original sound, and the decoded waveform has a large amplitude change that is not present in the original sound waveform.

[Means to solve the problem]

The ADPCM decoder of the present invention converts a digital signal obtained by sequentially accumulating difference values selected by the values of both ADPCM code data and a quantization width pointer into an analog signal, and outputs the analog signal. In an ADPCM decoder that decodes data and sequentially updates the value of a predetermined quantization width pointer, there is a register that accumulates and holds the difference value up to the current sampling point, and a register that stores the cumulative difference value up to the current sampling point. , an addition/subtraction operation that outputs a calculation output that includes addition or subtraction of the difference value of the next sampling point, and also outputs a control signal that is involved in generating a carry signal or borrow signal that occurs in response to a specific value of the addition or subtraction value. and receiving the control signal,
and means for controlling whether or not the register can take in the calculation output of the adder/subtractor.

〔Example〕

Next, the present invention will be explained with reference to the drawings. FIG. 1 is a block diagram of one embodiment of the present invention. As shown in FIG. 1, this embodiment includes an L register 1, a decoder 2, a limiter 3, an A register 4, a difference value data table ROM 5, an adder/subtractor 6, an X register 7, and a D
It is configured to include an A converter 8, an adder/subtracter 9, and an AND circuit 10.

Moreover, FIG. 2 is a waveform diagram of the original sound waveform encoded data, and FIG. 3 is a waveform diagram when the encoded data shown in FIG. 2 is decoded according to the embodiment of the present invention. .

In FIG. 1, the outline of the operation of the circuit system including the L register 1, decoder 2, limiter 3, A register 4, difference value data table ROM 5, adder/subtractor 6, DA converter 8, and adder/subtractor 9 is as follows. This is the same as in the conventional example described above. The difference from the conventional example of the present invention is that the function of the adder/subtractor 6 is different, and the latch signal 107 input to the X register 7 is changed from the level signal 107 output from the adder/subtractor
6 and is gate-controlled in an AND circuit 10.

In FIG. 1, AD latched by L register 1
PCM code data 101 is input to both the decoder 2 and the difference value data table ROM 5. In the decoder 2, the amount of movement of the quantization width pointer is determined by decoding the ADPCM code data 101. The output of the decoder 2 is added to or subtracted from the output of the A register 4 which latches the value of the quantization width pointer in the adder/subtractor 9, and the result of the addition/subtraction is input to the limiter 3 and the value of the quantization width pointer is It is limited within the upper and lower limits and is input to the A register 4.

In the difference value data table ROM 5, an address is selected by the input values from the L register 1 and the A register 4, and amplitude data 103 and code data 104 of the difference value of the next sampling pointer are output. These amplitude data 103 and code data 104 are input to an adder/subtractor 6, where addition and subtraction processing is performed with the cumulative result of the difference value input from the X register 7, and the amplitude value of the next sampling point is calculated. A gate control signal 106 generated by inverting a carry signal or a borrow signal is output as well as a calculation output 105.

This calculation output 105 is input to the X register 7, and the gate control signal 106 is sent to the AND circuit 10. A latch signal 107 for the X register 7 is also input to the AND circuit 10, and the latch signal 107 is input to the X register 7 only when the gate control signal 106 is at a high level.

The digital output of the X register 7 is DA-converted by a DA converter 8 and output as an analog signal 108.

When sampling and ADPCM encoding the large-amplitude original sound waveform shown in Fig. 2, after obtaining the sampling value S2 by adding the difference value Δ2 to the sampling value S1, the next difference value is the original sound waveform. If Δ3 with the minimum error is selected and ADPCM encoded, S
The amplitude of 3 exceeds the maximum value of FFH.

On the other hand, after subtracting the difference value Δ8 from the sampling value S7 to obtain the sampling value S8, if Δ, which minimizes the error from the original sound waveform, is selected as the next difference value and ADPCM encoding is performed, the difference value is The accumulated amplitude of S is less than or equal to the minimum value of 0.

When the ADPCM code data obtained by encoding the amplitude data from Sl to S12 shown in FIG. 2 is reproduced by the decoder according to the embodiment of the present invention shown in FIG. 1, as shown in FIG. The value obtained by adding Δ2 to X2 becomes X2, and since the added value exceeds FFH when Δ is added to becomes low level. In this case, as mentioned above, AND
Latch signal 107 is gated off in circuit 10;
It is not input to the X register 7. Therefore, the addition output that exceeds FFH in the adder/subtractor 6 is
is not latched, and the above-mentioned X is held at the same amplitude value as x2. Further, even when Δ4 is subtracted from Xl, since no borrow signal is generated in the adder/subtractor 6, the gate control signal 106 becomes high level. Therefore, the latch signal 107 is input to the X register 7 via the AND circuit 10, and the normal operation in which the subtraction result X4 in the adder/subtractor 6 is latched into the X register 7 is performed.

Next, the value obtained by subtracting Δ8 from X7 becomes X8, and since the subtracted value becomes less than O when Δ9 is subtracted from x8, a borrow signal is generated in the adder/subtractor 6, and a gate control signal is output from the adder/subtractor 6. 106 becomes low level. When the gate control signal 106 becomes low level, the latch signal 107 is gated off by the AND circuit 10 and is not input to the X register 7. Therefore, when the subtraction output in the adder/subtractor 6 becomes 0 or less, the subtraction output is X
It is not latched by register 7, and X9 is held at the same amplitude value as X8. Further, even when Δto is added to X9, a carry signal is not generated, so the gate control signal 106 becomes high level, and the addition result x0° of the adder/subtractor 6 is latched into the X register via the latch signal 107. Normal operation is performed.

[Effects of the Invention] As explained above, the present invention adds or subtracts the difference value of the next sampling point in the process of decoding an ADPCM code, corresponding to the X register that latches the result of sequentially accumulating the difference values. In the adder/subtracter that calculates the amplitude value, when a carry signal or a borrow signal is generated according to the result of the addition/subtraction, the latch signal to the X register is stopped, and the X register stops taking in the result of the addition/subtraction. This has the effect of eliminating the defect that amplitude changes that do not exist in the original sound waveform occur in the decoded waveform of the ADPCM code.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing an embodiment of the present invention, FIG. 2 is a waveform diagram of original sound waveform encoded data corresponding to the embodiment, and FIG. 3 is composite data during decoding according to the embodiment. 4 is a block diagram showing a conventional example, and FIG. 5 is a block diagram showing a conventional example.
a waveform diagram of original sound waveform encoded data corresponding to the conventional example;
FIG. 6 is a diagram showing composite data upon decoding according to the conventional example. In the figure, 1... L register, 2...
・Decoder, 3...Limiter, 4...
A register, 5...Difference value data table RO
M, 6, 9...Adder/subtractor, 7...X register, 8...DA converter, IO...
AND circuit.

Claims (1)

[Claims] A digital signal obtained by sequentially accumulating difference values selected by the values of both ADPCM code data and a quantization width pointer is converted into an analog signal and outputted, and the ADPCM code data is In an ADPCM decoder that decodes and sequentially updates the value of a predetermined quantization width pointer, there is a register that accumulates and holds the difference value up to the current sampling point, and the following for the register: an adder/subtractor that outputs a calculation output including addition or subtraction of a difference value between sampling points, and outputs a control signal involved in generating a carry signal or borrow signal that occurs in response to a specific value of the addition value or subtraction value; An ADPCM decoder comprising at least the following: , means for receiving the control signal and controlling whether or not the register can take in the calculation output of the adder/subtractor.