JP2856769B2 - Speech synthesizer - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Industrial application field) The present invention relates to a speech synthesizer that can naturally synthesize a coded numeric string with speech.

(Prior Art) In general, when a rule-based speech synthesizer outputs a digit string, for example, "3561", when it represents a price, it is read as "Sanzen / Gohjak / Rokju / Ichien". In the case of a telephone number, it is preferable not to read digits such as “Sun / Go / Roku / Ichi”.
Conventionally, if you do not do this reading, Insert an accent-like delimiter for each single digit like
Pauses were inserted at accent breaks to make it easier to hear.

However, when a person actually utters a number, utterance is rarely performed while separating each sound of a number, and utterance is often performed continuously.

(Problems to be Solved by the Invention) As described above, when outputting a digit string coded by a conventional speech synthesizer without digitizing, it utters one digit of a digit, which is unnatural. Had become.

Therefore, the present invention provides a speech synthesizer that can synthesize a character string such as a telephone number with a natural sound by using an input number string as one accent phrase every two digits from the beginning. The purpose is to provide.

[Structure of the Invention] (Means for Solving the Problems) A speech synthesizing apparatus according to the present invention, which outputs speech without digit reading of a coded given number sequence, includes two digits from the beginning of an input number sequence. Prosody parameter generation means for generating a prosody parameter with each digit as an accent boundary, and means for generating and outputting a synthesized sound based on the prosody parameter generated by the prosody parameter generation means.

(Operation) According to the speech synthesizer of the present invention, every two digits from the beginning of a coded number string are used as accent boundaries to accentuate a two-digit number or the remaining one-digit number. Can be generated with good naturalness.

Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

In FIG. 1, reference numeral 1 denotes an input unit for inputting a coded character string and a numeric string. Reference numeral 2 denotes an accent testing unit for comparing a code given through the input unit 1 with an accent dictionary 3 to obtain an accent type for the input,
Read and get part of speech information. Here, when the part-of-speech information corresponding to the input is “number” representing a number, two digits of the number constitute one accent phrase.

The accent dictionary 3 contains accent types for words,
Readings, parts of speech, etc. are registered. The synthesis parameter generation unit 4 generates a prosody parameter based on the accent type given from the accent test unit 2 and generates a phoneme parameter by referring to the speech unit file 5 based on the reading. Send to 6. The speech unit file 5 stores parameters obtained by analyzing each syllable.

The synthesizing unit 6 generates a synthesized sound based on the prosodic parameters and the phonological parameters sent from the synthesis parameter generating unit 4, and outputs it.

Next, in such a configuration, for example, “75387”
A case will be described in which a numeric string is input. The numeral string “75387” is sent to the accent testing unit 2 via the input unit 1. In the accent test unit 2, by comparing with an accent dictionary 3 as shown in FIG.
All parts of speech are divided into "75", "38" and "7" because of the "number". At this time, the accent-type reading is performed using the digit-combined accent information of “5”. (Type 3), using the digit-combined accent information of "8" Is tested. This information is sent to the synthesis parameter generation unit 4 and refers to the speech unit file 5. Are generated and sent to the synthesis unit 6, respectively. The synthesizer 6 generates a synthesized sound based on the synthesis parameters and outputs the sound.

In addition, the number reading is all 2 mora ("2" is "knee"
"5" is "go"), so two numbers are 1
If the accent phrase is 4 moras, and all the digit combination accent information is “1”, any combination of accent types will be “type 3”.

If one number remains, the accent type is set to “1”. These rules can be implemented, for example, by the numeric accent testing unit 7 shown by a broken line in FIG.
In this case, the input unit 1 detects a number and sends it to the number accent testing unit 7. The numeric accent detector 7 divides the numeric string into two digits from the beginning, and uses the two-digit number as one accent phrase. Then, the accent type of the two-digit number is set to “3”, and if one digit remains, the accent type of the number is set to “1”, and the reading of the number and the accent type are sent to the synthesis parameter generation unit 4. send.

As described above, when a character string such as a telephone number is output as speech, by setting every two digits from the beginning of the number string as an accent boundary, natural utterance of the number string can be achieved.

The present invention is not limited to the embodiments described above. For example, the schematic configuration diagram of the speech synthesizer shown in FIG. 1 is merely an example, and the contents of the accent dictionary in FIG. 2 are not limited to the above-described embodiment. Also, the part of speech that represents a number need not be “number”. In short, the present invention can be implemented with various modifications without departing from the scope of the invention.

[Effects of the Invention] As described in detail above, according to the present invention, every two digits from the beginning of an input numeric string are set as accent boundaries, and two digits or the remaining one digit is accented. Accordingly, it is possible to provide a voice synthesizing apparatus capable of obtaining a natural voice output similar to a voice uttered by a human.

[Brief description of the drawings]

FIG. 1 is a schematic configuration diagram of a speech synthesizer according to one embodiment of the present invention, and FIG. 2 is a diagram showing the contents of an accent dictionary in the embodiment. 1. Input unit 2. Accent test unit 3. Accent dictionary 4. Synthetic parameter generating unit 5. Speech unit file 6. Synthetic unit 7. Numerical accent test unit.

Claims (2)

(57) [Claims] 1. A prosody parameter generating means for generating a prosody parameter with an accent boundary every two digits from the beginning of an input digit sequence, wherein the encoded digit sequence is output as voice without digit reading. Means for generating and outputting a synthesized speech based on the prosody parameters generated by the prosody parameter generation means. 2. A speech synthesizer according to claim 1, wherein said prosody parameter generation means generates a prosody parameter by setting the accent type divided every two digits to "type 3".