JPS63231500A - Voice encoding 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 (Field of Industrial Application) The present invention relates to a voice encoding system using a delta modulation method used in digital mobile radio 2 communications and the like.

(Prior art) In general, as a method for encoding an audio signal, the audio signal is divided into certain intervals and sequentially sampled, and for each sample, the difference between the sample value and the sample value one sample before is extracted.
A coding method using a delta modulation method is known in which an audio signal is transmitted by quantizing the difference in one bit.

A conventional speech encoding method will be explained with reference to FIG.

Figure 5 is a diagram showing the encoding and decoding of audio using the conventional audio encoding method using the delta modulation method.
is a diagram showing a sequence of sample values obtained by dividing the input audio signal into certain intervals, and Figure 5 (B) is a diagram showing the difference between each sample and the sample value of the previous sample, and the code sequence is a diagram that shows the difference signal between the two samples. If the difference is positive, it is expressed as 1, and if it is negative, it is expressed as O. Figure 5 (C) is based on the code string shown in Figure 5 (B), and if the sign is 1, it is expressed as +Δ and the previous sample value is expressed. Δ is added to the sign of 0.
In this case, the encoded speech is decoded by subtracting Δ from the previous sample value as −Δ. Δ is a step size whose value is determined by the variance value of the amplitude of the input audio signal. Usually, such a coding method is called a coding method using the delta modulation method, but in this way, the conventional Voice can also be easily encoded using the voice encoding method.

(Problems to be Solved by the Invention) However, the conventional delta modulation method described above has the following problems.

(1) - Since a 1-bit code is assigned to each sample value, the number of bits that can represent one second of audio signal (hereinafter referred to as bit rate) is determined by the sampling frequency. It is not possible to compress the amount of information any further.

(2) Even though the waveform of the original voice hardly changes, if the sign of the difference signal is "0, 1, 0, 1...", the decoded signal will change step-like by the value of the step size. So-called granular noise, which is caused by repeated changes, occurs.

An object of the present invention is to provide a speech encoding method using a delta modulation method that can compress the amount of information with a simple configuration.

(Means for Solving the Problems) In order to solve the above problems, the present invention provides an audio encoding system that uses a delta modulation method, which is one of the audio encoding methods, in which the input audio signal is kept constant. Divide into sections, and make each section silent or silent.

Voicedness is determined, and based on the determination, no encoding is performed in the silent section, encoding is performed using a normal delta modulation method in the voiced section, and a delta modulation method is applied for each sampled value in the voiced section. By performing encoding using , the overall bit rate can be lowered and noise can be prevented.

(Function) With the above-described configuration, the present invention divides an input audio signal into fixed sections, and performs silent and silent processing for each fixed section.

If it is determined that there is voicing, and if it is determined that there is no voice, there is no need to allocate bits because no encoding is performed.If it is unvoiced, encoding is performed using the normal delta modulation method to express the signal fluctuations as they are. In the case of sound, multiple sample values are regarded as a stationary interval and encoded as a single sample value, so small fluctuations in the signal can be absorbed and bit allocation can be reduced, so granular noise can be prevented.
You can also lower the overall bitrate.

(Example) An example of the present invention will be described with reference to FIGS. 1 to 4. FIG. 1 is a block diagram showing the configuration on the encoding side of a speech encoding system according to an embodiment of the present invention, FIG. 2 is a block diagram showing the configuration on the decoding side of the same embodiment, and FIG. 3 is a block diagram showing the configuration on the decoding side of the same embodiment. FIG. 4 is a schematic flowchart showing the encoding method of the same embodiment. In the same figure, 1
is an A/D converter, 2 is an unvoiced/voiced determiner, 3 is an encoder,
4 is a decoder, 5 is a D/A converter, and n is a sequence of sections obtained by dividing the audio signal into constant time intervals.

The A/D converter 1 divides the input audio signal into certain sections based on the sampling frequency and samples the signals. The voiceless/voiced determiner 2 detects silence for each sampled section.

A determination is made as to whether the segment is voiced or not, and the encoder 3 creates a coded string based on the determination made by the voiced/voiced determiner 2 for each section. The decoder 4 performs decoding based on the determination for each section and the encoded sequence sent from the encoding side, and the D/A converter 5 reproduces it.

Next, the operation of the above embodiment will be explained. The audio signal shown in FIG. 3 input to the A/D converter 1 is sampled at regular intervals based on the sampling frequency. The sampled signals are sequentially sent from the A/D converter 1 to the unvoiced/voiced determiner 2, and the sampled signal strings are stored every fixed interval of about several tens of m5ec. The unvoiced/voiced determination base 2 determines voiced, unvoiced, or silent for each section for this section, and the decoder 3
The results of the judgment will be handed over to the Therefore, until the section length data is stored in the unvoiced/voiced determiner 2.

This method results in a time delay.

On the other hand, the signal sampled by the A/D converter 1 is silent/
At the same time as being sent to the voicing determiner 2, it is also sent to the encoder 3. The encoder 3 immediately starts encoding when the determination result of the voiceless/voiced determiner 2 is input.

A specific encoding method will be explained with reference to the flowchart of FIG. If section 1 of the section order sequence n in FIG. 3 is determined to be silent, only the silent frame bits elapse as shown in the flowchart, and no coded string data is sent. Therefore, the number of bits in this section is only silent frame bits.

If section 3 of section sequence n shown in FIG. 3 is determined to be unvoiced, an unvoiced frame bit is set and conventional delta modulation coding is performed. Since one bit is one sample in this section, the number of bits in the section is the sum of the number of samples and the number of unvoiced frame bits.

When section 6 of the section sequence n shown in FIG. 3 is determined to be voiced, voiced frame bits are set and delta modulation is performed, for example, every two samples. In this case, since it is only necessary to check the value once every two samples, the number of bits is the value obtained by adding the number of voiced frame bits to the value of the number of samples/2.

The above is an example of the present invention, and the step size is fixed, but the bit allocation is reduced, so if you prepare about four step sizes and send information about which step size to use, good.

To select the step size in this case, when the unvoiced/voiced determiner 2 determines whether the voice is unvoiced or voiced, an amplitude variance value is obtained, and the step size is determined based on that value. Therefore, when changing the step size, the amount of information increases accordingly, so the bit allocation increases by several bits per section and the bit rate increases, but changes in the signal can be expressed more precisely and overload noise can be prevented.

(Effects of the Invention) According to the present invention, in the conventional delta modulation method, the coding method is determined for each section by determining silence, p, voice, and voicing for each short section, so that overall bit allocation is reduced. This has the effect of reducing the bit rate and preventing so-called granular noise.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration on the encoding side of a speech encoding system according to an embodiment of the present invention, FIG. 2 is a block diagram showing the configuration on the decoding side of the same embodiment, and FIG. 3 is a block diagram showing the configuration on the decoding side of the same embodiment. FIG. 4 is a schematic flow diagram showing the encoding method of the same embodiment, and FIG. 5 is a speech code obtained by the conventional delta modulation method. It is a figure which shows an example of decoding. 1... A/D converter, 2... Voiceless/voiced determiner,
3... Encoder, 4... Decoder, 5... D/
A converter, n... An ordered sequence of sections in which the audio signal is divided into fixed time intervals. Patent applicant: Matsushita Electric Industrial Co., Ltd. Figure 1 Figure 2 Meiyo [Figure 4?

Claims (1)

[Claims] In an audio encoding method using a delta modulation method, an input audio signal is divided into predetermined sections, and a determination is made as to whether each of the predetermined sections is silent, unvoiced, or voiced, and based on the determination, encoding is performed in the silent section. First, the unvoiced section is encoded using a normal delta modulation method, and the voiced section is encoded using a delta modulation method for each of the plurality of samples, thereby lowering the overall bit rate and preventing noise. A speech encoding method characterized by: