JPS5887595A - Voice accumulation coding system - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention uses an encoding method K11l when an audio analog signal is encoded and stored in a message.

Conventionally, this type of audio storage encoding method has two methods: (1) a method that encodes and stores the input signal regardless of whether it is a speech section or a silent section; However, in a voice section, the input signal is encoded and stored, but in a silent section, the length of time of the section or the start and end times of the section are encoded and stored using a linear function.

However, in the former method (1), the storage efficiency is extremely poor because the storage memory is used during silent periods, which can be long depending on the case.On the other hand, in the latter method (2),
), but the time length and the number of bits required to encode the company time will increase to represent a very wide range of silent intervals, so it is equivalent to amount of memory is required.

SUMMARY OF THE INVENTION In view of the above-mentioned drawbacks of the prior art, it is an object of the present invention to provide a voice storage encoding method that can improve the % product efficiency without impairing the quality of stored voice. Characteristics, silent interval t-
Although it is necessary to express short silent sections with high precision, it is sufficient to express relatively low precision for long silent sections. However, based on the knowledge that it is acceptable to express this as a fixed period of time, the length of the silent section is encoded using a nonlinear encoding function that reduces the accuracy of the expression as it becomes longer. Hereinafter, the present invention will be explained based on the drawings. 0 jllaQ shows a simple example of a nonlinear encoding function,
The horizontal axis represents the actual time length of the silent section, and the vertical axis represents the quantization step. In this example, the time length of the silent section is divided into four points: i, 2t, 4, and 8t (where t is a constant value). , 4t to 8t4D, and 8 are classified into nonlinear S steps D1 to Ds. That is, in a short silent section, the steps are close together, and as the silent section becomes longer, the steps become more sparse.Furthermore, all steps above a certain period of time 8 are the same step.

When a non-linear encoding function as shown in Fig. 1 is used, the length of the silent section can be determined using a small amount of information Dl-Di, I! Therefore, the encoded signal indicating the time length of the silent section has a low bit value, and the capacity of the memory for storing it can be reduced. of course,
Since the non-linear encoding function takes into consideration the properties of the silent section described above, the quality of the stored speech is not impaired.

jlI2 diagram shows one embodiment of a device for realizing the present invention.

In the figure, a voice discrimination circuit 1 is a known circuit that discriminates between a voice section and a silent section with respect to an input voice signal Min.
The input audio signal Vin is output separately into a signal S■ for the audio section and a signal SS for the silent section. Audio encoding circuit 2
is a known circuit, and the voice section signal SVtPCM+A
It encodes using a suitable audio encoding method such as DPCM, and outputs the encoded signal DV. The silent section length measuring circuit 3 is a circuit that receives the signal SS of the silent section, measures the actual time length of the silent section, and outputs a signal indicating the value, that is, a silent section length signal TSk.

The nonlinear conversion circuit 4 is a circuit that inputs the silent interval length signal TS and outputs an encoded signal DB of the corresponding quantization step based on the nonlinear encoding function shown in FIG. 1. Signal DV of the voice section encoded in this way
and the silent interval length signal D8) are combined by the code editing circuit 5), edited in the same order as the input sound 11yVin, and outputted as the coded code Vout for accumulation.

Now, considering the input audio signal vin shown in j13all(-), the audio section V1 for this input audio signal Vin
m Vg and the silent interval at -Sm are determined for each voice interval Vl, in which the signal within that interval is encoded, 1 for each silent interval 81e and 8 for the interval O time length is measured. The value is encoded through non-linear transformation. As shown in FIG. 3(b), encoded information DVs corresponding to the voice section Vl, encoded information D corresponding to the voice section Sl.
os, the encoded information Dv2 corresponding to the voice section, and the encoded information Dam corresponding to the silent section SxK are edited into the storage encoded signal Vout.

The above-mentioned storage encoded signal Vout is stored in the memory as before, and then read out for information transmission and information m1lK#.
be done. In addition, in the case of demodulation to an analog signal, DSx
According to the present invention, as explained above, according to the present invention, the silent interval is encoded using a non-linear encoding function in which the accuracy of the expression decreases as the time length of the silent interval increases. , storage efficiency is improved without sacrificing audio quality.

[Brief explanation of the drawing]

FIGS. 1 to 3 relate to the present invention; FIG. 1 is a graph showing an example of non-linear encoding, FIG. 2 is a block diagram showing an example of an audio storage encoding board, and FIG.载) is an operation explanatory diagram. In the drawing, ■in is the input audio signal, vout is the coded signal for storage, l is the audio discrimination circuit, 2 is the audio encoding circuit, 3 is the silent interval length measuring circuit, 4 is the nonlinear conversion circuit, and 5 is the code editing circuit. , v yo and heptad are voice sections, and St and Sm are silent sections. Patent Applicant: International Telegraph and Telephone Co., Ltd. Patent Attorney Patent Attorney Shibu Mitsuishi (and 1 other person) Figure 1 51, Hi-ku Ken no Harumun Restaurant Figure 2

Claims (1)

[Claims] a speech discrimination circuit that discriminates an input speech signal into a speech section and a silent section; a speech encoding circuit that encodes the signal of the speech section; a silent section length column setting circuit that sets the time length of the silent section; A nonlinear transformation is performed to encode this fixed time length using a nonlinear encoding function, which reduces the accuracy of the representation as it becomes longer.
A code edition #l that collects the audio encoded signal of the audio section and the encoded signal of the silent section length in the same order as the input audio signal and outputs the encoded signal for storage. - A voice storage encoding method consisting of .