JPS59102297A - Voice synthesizer - 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 relates to a speech synthesis device, and more particularly to a synthesis device having a circuit for interpolating two sets of temporally adjacent speech parameters.

Conventionally, the spectral envelope parameter was calculated using natural speech rh momme (
In order to obtain -a, l-to J-r^711-+-,,,-,-, there is a known device that uses uniform interpolation over a period shorter than the previous 6α minutes @period. .

However, in the above-mentioned device, the same method is used for interpolation regardless of whether the spectral envelope parameter changes sharply or remains unchanged, so the synthesized speech becomes unnatural compared to natural sounds. There were drawbacks.

The purpose of the present invention is to perform an extrapolation operation to
An object of the present invention is to provide a speech synthesis device that outputs synthesized speech close to natural speech.

The present invention provides a speech synthesis device equipped with a circuit for interpolating speech parameters between temporally distinct frames, in which the interpolation circuit has an interpolation circuit that interpolates speech parameters between temporally distinct frames.
The present invention is characterized in that it includes a plurality of functions such as a non-interpolation function, and has means for selecting a predetermined later layer from among the interpolation circuits according to differences in sound F parameters of multiple matching frames.

According to this invention, it is possible to follow the transient changes in the sound of the material, which are subject to rapid transient changes, and if the envelope parameters of the preceding frames are not different from each other depending on the difference between the two, The waveform can be smoothed by comparing the value with a pre-added value and selecting either linear interpolation or non-linear interpolation depending on the result. Therefore,
Whether using parameters extracted from natural speech or artificially set parameters, synthesis can be performed with excellent sound quality.Examples of the present invention are described below with reference to the drawings. I glared at him. FIG. 1 is a block diagram showing an embodiment of the present invention. ip synthesis communication 1, spectrum color glaze parameters used for synthesis ↑PI
R (J
A composite information storage circuit 2 such as M or RAII; There is a non-interpolation circuit 4 which directly outputs the input synthesis information from the synthesis information storage circuit 2 to the output terminal 13 without subjecting it to the synthesis information input from the synthesis information storage circuit 2. A nonlinear interpolation circuit 5 performs nonlinear interpolation ti on the synthesized information and outputs it to an output terminal 14, and a nonlinear interpolation circuit 5 performs linear interpolation on the synthesized information inputted from the synthesis information storage circuit 2 and outputs it to an output terminal 15. The output linear interpolation circuit 6 and the synthesis information recorder detect the amount of change between each smooth tram envelope parameter information of the temporally irregular frames output from the circuit 2, and detect the temporal irregularities. a detection circuit 8 that detects a change in voiced/unvoiced discrimination information and quickly selects a different output terminal of the interpolation circuit according to the change; , a synthesis filter 8 that performs voice synthesis by setting the filter coefficients, the pitch, voiced/unvoiced discrimination information, and converting the synthesis output from the synthesis filter 8 into an analog signal 1)/ Reading of synthesis information from the A change change circuit 9, synthesis information storage circuit 2, interpolation circuit 3, detection circuit 7, synthesis filter 8. J) A control circuit 10 that controls each operation of the /AA conversion circuit 9, a command input terminal 11 that inputs commands to start and stop voice synthesis operations to the control circuit 10, and an output that outputs synthesized voice. All terminals 12 are connected internally.

The speech synthesis device 1 activates the control circuit 10 in accordance with a command input from the command input terminal 11 and starts synthesis processing. The control circuit 1o provides the synthesis information storage circuit 2 with addressing of spectrum envelope parameter emotion, voiced/unvoiced discrimination information, pitch information, sound source amplitude information, temporal and
Interpolation circuit 3, with the combined information @whole bladder
Input to detection circuit 7. A non-interpolation circuit 4, a non-linear interpolation circuit 5, and a linear interpolation circuit 6 inside the interpolation circuit 3 perform predetermined interpolation on the composite information, and then output the output terminal 13.1.
Output on 4.15. The detection circuit 7 detects whether the synthesized information inputted from the synthesized information storage circuit 2 has different pieces of voiced/unvoiced discrimination information that are different in terms of time, and when there is a change from unvoiced to voiced, the synthesized information The output terminal 13 holding the synthesis information of the previous frame is selected and inputted to the synthesis filter 8. On the other hand, if there is no change in the voice/unvoice discrimination information that has been present at different times in the synthesis information inputted from the synthesis information storage circuit 2, or even if there is a change, M?
In the case of a change from to silent, the output terminal 13 is not selected, but the amount of change in the spectrum envelope parameter is detected, and compared with a threshold value preset in the detection circuit 7, the amount of change is determined. The output terminal 14 of the nonlinear interpolation circuit 5 is selected for the threshold value, and the output terminal 14 of the nonlinear interpolation circuit 5 is selected.
is smaller than the set threshold, the output terminal 15 of the linear interpolation circuit 6 is selected and the interpolated value is output to the synthesis filter 8.

As described above, according to this embodiment, no interpolation, linear interpolation,
The results of nonlinear interpolation can be selectively used, and speech close to natural speech can be synthesized. Furthermore, it is clear from the drawings that the hardware mechanism is extremely simple and can be easily created using well-known semiconductor technology.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing one embodiment of the present invention. DESCRIPTION OF SYMBOLS 1... Speech synthesis device, 2... Synthesis information storage circuit, 3... Interpolation circuit, 4... Non-interpolation circuit section, 5... ...Nonlinear interpolation circuit section, 6...
...Linear interpolation circuit section, 7...Detection circuit, 8
...Synthesizing filter, 9...D/A fan, 10...Control circuit, 11...Command input terminal, 12...Output Terminal, 13...
Musashi output y-child, 14...Non-wagata sleeve +i4]
Output terminal, 15... glandular body. Output terminal.

Claims (1)

[Claims] In a speech synthesis device having a circuit for interpolating audio parameters between temporally adjacent frames, the interpolation circuit has a function of interpolating different types of Be, and the difference between the audio parameters of adjacent frames. A voice synthesizer that can select any & function according to the situation.