JPS6146997A - Voice reproduction 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 [Technical Field 1] The present invention relates to a voice reproduction method that can arbitrarily change the speaking speed.

[Background Art 1 Calculators have been developed that notify the contents of recently pressed keys using synthesized sounds. When using this to input programs into a program calculator, personal computer, etc., it is better to say the inputted content slowly while the keys are manually inputting, and then to say it quickly again when the input is finished and you want to confirm the content. It's convenient. In order to make the same word uttered at both normal speed and fast speed, it is possible to have a synthesized sound that is analyzed and synthesized from two types of original speech, normal speed and fast speed. The disadvantage is that the storage capacity becomes large. Another method is to change the sampling frequency, but in that case, the disadvantage is that the reproduced synthesized sound will be very unnatural.

[Objective of the Invention 1 The present invention has been devised in view of the above-mentioned problems, and its purpose is to synthesize speech at normal speed and rapid speech by simply adding a few circuits to the conventional speech synthesis circuit. To provide a voice reproduction method capable of outputting two types of synthesized sounds.

[Disclosure of the Invention 1] However, the audio reproduction method of the present invention performs analysis using the original voice uttered rapidly, and uses the parts where steady parts such as vowels or unvoiced sections continue for a certain period of time as repeat part data. The feature is that the synthesized sound is extended by adding it to the voice parameters and storing it, and repeating the synthesized voice at the repeat point a predetermined number of times during synthesis.I) The consonant part of the voice is By stretching the vowel part in the time axis direction without changing it, the speech rate can be changed without impairing naturalness.

FIG. 1 shows an example in which the method of the present invention is implemented in an analysis-synthesis type speech synthesis device that uses a limited number of words. In the figure, the memory stores voice parameters and repeat location data for each word, and these data a are sequentially input to the control circuit C during synthesis. A speaking rate instruction signal is inputted to the control circuit C from the speaking rate setting section, and input data a is selected and only the normal voice parameter C is sent to the voice synthesis circuit S. However, when the repeat point is reached, The audio parameters at this repeat location are repeatedly output as many times as instructed by the utterance speed instruction signal. In this way, the synthesized sound d output from the speech synthesis circuit S is played back in a more expanded manner than the original sound.

The embodiment shown in FIG. 2 includes a random number generator R in the control circuit C,
During the repeat period, a random number r is added to the pitch parameter p, which is one of the audio parameters, and the result is output. With this configuration, it is possible to prevent repeat sounds from becoming monotonous and to impart naturalness to the synthesized sound. For example, the average pitch period of a synthesized sound is 5m5 (200Hz)
If this is to be varied by 10%, the output of the random number generator R should be generated between ±0.5n+S. Note that by applying the same processing to the amplitude parameter as described above, the naturalness of the reproduced sound can be further increased.

The above two examples use a so-called analysis-synthesis type speech synthesis circuit, but this method has the disadvantage that the synthesis circuit becomes complex. Therefore, in order to simplify the circuit, it may be possible to use a waveform encoding method such as the PCM method or the D PCM method, but in that case, there are problems such as the following6. Because of this, the synthesized sound changes smoothly even if the parameters change discontinuously, but with waveform encoding, if you try to directly connect the waveforms, discontinuities will occur and the synthesized sound will not change smoothly.

Therefore, the method of repeating as in the present invention does not work as it is. The embodiment shown in FIG. 3 solves this problem, and first, as in the case of FIG. 1, an original voice uttered rapidly is used and analyzed by an analysis and synthesis system.

Next, the stationary portion is extracted and its parameters are repeated three times to synthesize the entire waveform. Next, the waveform encoding system treats this waveform as the original audio and encodes it, extracts and encodes the central part of the repeat part, and adds this repeat waveform code to the entire waveform code together with the repeat part data. As shown in Fig. 3, when synthesizing 1 to be memorized, the control circuit C selects these data e and outputs the normal waveform code f as it is, and when the repeat point is reached, the data is repeated a predetermined number of times according to the utterance speed instruction signal. It outputs the waveform code repeatedly. With this configuration, the analysis system becomes slightly more complex, but the circuit configuration of the synthesis system becomes extremely simple, especially for applications such as program calculators and personal computer input devices that use a limited number of words and have a small storage capacity. This has the advantage that the system can be provided at low cost.

[Effects of the Invention] As described above, the present invention analyzes the original speech uttered rapidly, extracts the stationary part, adds it to the speech parameters as repeat part data, and adds the repeat part data to the speech parameters during synthesis. It repeats the synthesized sound and extends the synthesized sound, so by selecting whether to repeat or not, it can be played at normal speed or fast, and it can also be played at normal speed or fast. Since there is no need to prepare regular parameters, there is no need to increase the memory capacity too much, and since the method does not change the number of sampling cycles, the pitch of the voice does not change and naturalness can be maintained. There is an advantage.

[Brief explanation of the drawing]

FIG. 1 is a block circuit diagram showing an embodiment of the method of the present invention;
FIG. 2 is a block circuit diagram showing an embodiment of the pond, and FIG. 3 is a block circuit diagram showing yet another embodiment. S is a speech synthesis circuit, C is a control circuit, and R is a random number generator. Agent Patent Attorney Shi 1) Chief 7 Figure 1 @ 2 Figure Xi' jlT4) 2

Claims (1)

[Claims] (1) Perform analysis using the original voice uttered rapidly,
A synthesized sound can be created by adding and storing parts where a steady part such as a vowel or a silent section continues for a certain period of time as repeat part data to the voice parameters, and repeating the synthesized sound at the repeat part a predetermined number of times during synthesis. An audio reproduction method characterized by expansion.