JP2581050B2 - Voice analysis and synthesis device - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a speech analysis / synthesis apparatus that analyzes and encodes speech and then decodes and synthesizes the speech. , And a device for analyzing and synthesizing with sound source information such as a residual signal.

[Conventional technology]

Conventionally, when voice transmission is performed in a digital transmission system,
In order to compress the amount of information or perform confidentiality, multi-pulse driven linear predictive coding (MPEC), residual driven vocoder (RELP), adaptive predictive coding (APC), etc.
2. Description of the Related Art An encoding method is used in which a speech is analyzed and quantized and transmitted at regular time intervals to phonetic information (spectral information) such as linear prediction coefficients and Percoll coefficients and sound source information such as a residual signal and multipulse information.

[Problems to be solved by the invention]

By the way, in the above-mentioned encoding method, in order to quantize an analysis result to a limited transmission rate (bit rate), a quantization bit number is assigned to phoneme information and sound source information. In order to quantize each information more efficiently, a quantization method is statistically determined.

For example, the K parameter (Percoll coefficient) used as phonological information basically has an absolute value of 1 or less,
The distribution of possible values depends on the characteristics of the input speech.

FIG. 3 shows the K parameter distribution of voiced speech. FIG. 4 shows the K parameter distribution of unvoiced sounds. Conventionally, quantization is performed after defining these distributions and defining a distribution range. However, as is clear from FIG. 4, the K-parameter distribution of no sound has a disadvantage that the distribution range is narrow and the redundancy is extremely high in the general distribution range.

In view of the above drawbacks, an object of the present invention is to individually consider the distribution range of K parameters for voiced and unvoiced sounds from primary to higher order, and to allocate the number of K parameter bits without deteriorating quantization accuracy. It is to provide a speech analysis / synthesis apparatus that reduces the number of voices.

[Means for solving the problem]

According to the present invention, an analyzer having means for extracting sound source information of an input voice, spectrum extracting means for extracting spectrum information of an input voice, and means for quantizing the sound source information and the spectrum information, And a synthesizing unit having means for receiving and dequantizing the sound source information and the spectrum information, wherein the spectrum extracting means of the analyzing unit includes a K parameter ranging from a first order to a higher order as the spectrum information. Extracting, the analyzing unit further includes: a detecting unit that detects voiced or unvoiced input voice; a quantization range control circuit that selects a quantization range of the K parameter according to a detection result of the detection unit; Means for quantizing the sound source information and the K parameter according to a selection result by the quantization range control circuit; Means for transmitting the information and the K parameter together with the selection result by the quantization range control circuit, and the synthesizing unit includes means for selecting an inverse quantization range based on the selection result. A speech synthesis analyzer that performs

〔Example〕

Next, embodiments of the present invention will be described with reference to the drawings.

 FIG. 1 shows an analysis unit according to one embodiment of the present invention.

The voice input from the voice input terminal 1 is input to a spectrum information analyzer 2, a voiceless detector 3, and a sound source information analyzer 4. The output of the spectrum information analyzer 2 is connected to a spectrum quantizer 6, and the output of the sound source information analyzer 4 is connected to a sound source quantizer 7. The output of the voiced and unvoiced detector 3 is
The signal is input to the quantization range control circuit 5, and its output is input to the sound source quantizer 7, the spectrum quantizer 6, and the multiplexing circuit 8 as a quantization range control signal. The multiplexing circuit 8 multiplexes the spectrum information, the sound source information, and the quantization range control signal,
Output to the signal output terminal 9.

 FIG. 2 shows a synthesizing unit according to one embodiment of the present invention.

Each information input from the signal input terminal 10 is separated by the separation circuit 11
Are output to the spectrum inverse quantizer 12, the inverse quantization range control circuit 13, and the sound source inverse quantizer 14.
The output of the inverse quantization range control circuit 13 is a spectrum inverse quantizer.
12 and input to the sound source inverse quantizer. The dequantized spectrum information and sound source information are input to the synthesis filter 15,
The output of the synthesis filter 15 is output to the audio output terminal 16 as a synthesized sound.

〔The invention's effect〕

As described above, according to the present invention, the number of allocated bits can be reduced without deteriorating the quantization accuracy by detecting the voiced apex of the input speech and limiting the quantization range of the spectrum information when the voice is absent. Therefore, the reduced bits can be used for sound source information and the like, and there is an effect that voice quality can be improved.

[Brief description of the drawings]

FIG. 1 is a block diagram of an analysis unit of the speech analysis / synthesis device of the present invention, FIG. 2 is a block diagram of a synthesis unit of the speech analysis / synthesis device of the present invention, and FIG. 3 is a distribution diagram showing a voiced K parameter distribution. , FIG. 4 is a distribution diagram showing an unvoiced sound K parameter distribution. 1 ... voice input terminal, 2 ... spectral information analyzer, 3 ...
Voiced and unvoiced detector, 4 …… Sound source information analyzer, 5 …… Quantization range control circuit, 6 …… Spectrum quantizer, 7 …… Sound source quantizer, 8 …… Multiplexer, 9 …… Signal output terminal , 10 ... Signal input terminal, 11 ... Separation circuit, 12 ... Spectrum inverse quantizer, 13
…… Inverse quantization range control circuit, 14 …… Sound source inverse quantizer, 15…
... Synthesis filter, 16 ... Audio output terminal.

Claims (1)

(57) [Claims] A means for extracting sound source information of an input voice; a spectrum extracting means for extracting spectrum information of an input voice;
And an analysis unit having means for quantizing the sound source information and the spectrum information, and a speech analysis and synthesis device having a synthesis unit having means for receiving the quantized sound source information and the spectrum information and performing inverse quantization. The spectrum extraction means of the analysis unit extracts K parameters ranging from primary to high order as the spectrum information, and the analysis unit further includes detection means for detecting voiced / unvoiced input voice;
A quantization range control circuit that selects a quantization range of the K parameter according to a detection result of the detection unit; and quantizes the sound source information and the K parameter according to a selection result by the quantization range control circuit. Means, and means for transmitting the quantized sound source information and the K parameter together with the selection result by the quantization range control circuit, wherein the synthesizing unit selects an inverse quantization range based on the selection result. A speech synthesis analyzer characterized by comprising means for performing.