JPH077280B2 - Multi-pulse position quantization method - Google Patents

Description

The present invention relates to a multi-pulse position quantization method in a multi-pulse encoding device.

[Prior Art] On the analysis side in a multi-pulse encoding device, a linear prediction coefficient equivalent to the vocal tract transmission characteristic from a voice signal sampled for each frame and a vocal cord vibration wave, etc. A plurality of pulses (multi-pulses) whose number is about 1/10 or less of the number of voice samples for one frame whose amplitude is free are generated, and these linear prediction coefficients and multi-pulses are transmitted to the synthesis side as characteristic parameters. In a multi-pulse coding device for high-efficiency speech transmission, an efficient multi-pulse quantization method is extremely important in addition to the linear prediction coefficient quantization.

However, the number of multi-pulses is extremely smaller than the total number of samples in the frame, and it is considered that a multi-pulse train is a kind of the prediction residual waveform and that all the samples of the prediction residual waveform are quantized. It will be assigned and not a good idea.

In the conventional multi-pulse coding apparatus, instead of quantizing the total samples corresponding to the prediction residual waveform of one frame, the interval between adjacent multi-pulses and the amplitude of the multi-pulses are quantized to improve the quantization efficiency. I was trying to improve.

[Problems to be solved by the invention]

However, the conventional method of simply quantizing the interval of the multi-pulse has the following drawbacks. In code transmission of usually about 9600 bps in a multi-pulse encoder, for example, a frame cycle of 20 mSEC (at 8 kHz sample
In the case of 160 samples), the number of pal is 14, and 4 bits are assigned to each pulse for quantization of the amplitude position.
If there are 14 multi-pulses cut out for every 160 samples of a fixed frame, there are no pulses in the remaining 146 time slots, that is, the time slot subject to pulse interval quantization, but the time position is free. With multi-pulse, even if the pulses are evenly spaced, 146/14 = 10.4 sample intervals are required, and considering the degree of freedom of the pulse interval, it is not possible to express the pulse position information at 4bit intervals 0-15. It is known empirically that it may happen. Therefore, when the pulse interval cannot be expressed, that is, when the interval between pulses is 16 or more, this problem is circumvented by setting a pulse with an amplitude of "0" in the middle, and the pulse originally generated is deleted in return. It was a method of performing quantization while keeping the total number constant. This method has a drawback that the number of multi-pulses is substantially reduced by the position quantization.

Therefore, the present invention provides a high-performance multi-pulse position quantization method that eliminates such drawbacks as much as possible.

[Means for solving problems]

The method of the present invention cuts out an audio signal at regular frame intervals,
A method for quantizing a pulse position in a multi-pulse encoding device that expresses this frame with a linear prediction coefficient, and a plurality of pulses with free time position and amplitude, that is, a multi-pulse, wherein the frame is divided into a plurality of small frames. A multi-palace position quantization method of dividing and expressing whether or not a pulse exists in these small frames by a binary code, and for a small frame in which a pulse exists, quantizing the pulse interval for each small frame is there.

〔Example〕

 The present invention will now be described in detail with reference to the drawings.

FIG. 2 is a block diagram showing a multi-pulse encoding device. Input audio signal input via input line 1 is A / D
It is supplied to the converter 10 and output to the multi-pulse analyzer 11.

The multi-pulse analyzer 11 calculates a linear prediction coefficient, a multi-pulse that is a plurality of pulses having free time positions and amplitudes from the sampled speech signal, and outputs the multi-pulses to the quantizer 12. Also the decoder
The multipulse and linear prediction coefficient decoded by 22 are input to the multipulse synthesizer 21. Multi-pulse synthesizer 21
The synthesized voice signal is output to the D / A converter 20 and the voice is reproduced.

FIG. 1 is a block diagram showing an embodiment of the present invention. The multi-pulse input through the input line 40 shown in FIG. 1 and the multi-pulse cut out from the multi-pulse analyzer 11 at regular frame intervals are supplied to the divider 30 and divided into, for example, 4 small frames each having 40 samples. It is output to the determiner 31.

The pulse presence / absence determiner 31 determines whether or not a pulse exists for each small frame, and outputs the result to the code generator 32. The small-frame multi-pulse in which the pulse exists is supplied to the pulse position measuring device 33. The code generator 32 converts the input determination result into 4-bit information and sends it to the quantizer 34. The pulse position measuring device 33 measures the pulse interval for each small frame and outputs it to the quantizer 34. The quantizer 34 quantizes the measured pulse intervals and outputs the quantized data and 4b
The pulse judgment information of it is multiplexed and output.

As described above, in the conventional method, all the signals were input to the pulse position measuring device 33 and quantized.
1, by providing the code generator 32, pulse position measuring device 3
Since the number of frames calculated in 3 is reduced, a more effective quantization method is realized.

For example, assume that no pulse exists in one small frame when divided into four small frames.
In the conventional method, in pulse width 4bit quantization, 2-3
It is necessary to set individual pulses of amplitude “0”. That is, only the quantization of the pulse position is (2-3) x 4 = (8-12)
Requires extra bits. Of course, a pulse of amplitude “0” also requires quantizer quantization. Usually this amplitude quantization is 3
It is carried out at ~ 5bit. Therefore, when there is no pulse in one small frame, (4 + 3) is used in the conventional method.
Extra bits of about 2 × 14 = (4 + 5) × 3 = 27 bits are required. On the other hand, with the method according to the invention, an increase of at most 4 bits is sufficient.

In this embodiment, 1-bit information (indicating whether or not a pulse exists) is assigned to each of a plurality of small frames that are set by dividing the analysis frame, but it is inevitably the same as the number of small frames. Does not require the number of bits. This is true when there is no pulse in every small frame, such as none when there is no pulse in all small frames, and there is a bias when expressing the presence or absence of a pulse in a small frame for each frame. By implementing the above, a more efficient method can be configured.

〔The invention's effect〕

As described above, the present invention divides a constant frame represented by a multipulse whose time position and amplitude are free into small frames, and represents with a binary signal whether or not a pulse exists in these small frames, and , For a small frame in which a pulse exists, by quantizing the pulse interval for each small frame, it is not necessary to set a pulse of amplitude “0” even when the analyzed pulse interval of the multi-pulse is wide, and the efficiency is improved. There is an effect that a good multi-pulse position quantization method can be realized.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of the present invention, and FIG. 2 is a block diagram showing a multi-pulse encoding device. 10 …… A / D converter, 11 …… Multi-pulse analyzer, 12 ……
Quantizer, 20 ... D / A converter, 21 ... Multi-pulse synthesizer, 22 ... Decoder, 30 ... Divider, 31 ... Pulse presence / absence determiner, 32 ... Code generator, 33 ... … Pulse position measuring instrument, 34
…… Quantizer.

Claims (1)

[Claims] 1. A pulse position quantum in a multi-pulse encoding device which extracts a voice signal for every fixed frame and expresses this frame by a linear prediction coefficient and a plurality of pulses whose time position and amplitude are free, that is, a multi-pulse. In the coding method, the frame is divided into a plurality of small frames, whether or not a pulse exists in these small frames is expressed by a binary code, and the pulse interval is quantized for each small frame in which the pulse exists. A multi-pulse position quantization method characterized by the above.