JPH037994A - Generating device for singing voice synthetic data - Google Patents

Abstract

PURPOSE:To obtain a singing voice of high quality by inputting a musical score and a singing voice, extracting respective data on a character string, pitch, strength, length, and synthesizing a voice. CONSTITUTION:A score input part 1 extracts character string data as phoneme information and pitch data as rhythm information from the input score. The phoneme information extracting means consisting of a song voice input part 2, a feature extraction part 3, a segmentation processing part 4, and a rhythm information extraction part 5 extracts strength data and length data as rhythm information from the input song voice. The character string data, pitch data, strength data, and length data which are thus extracted are encoded by an encoding part 6 and outputted as the voice data of a voice synthesizing circuit X. Thus, the strength and length data are extracted according to an actual singing voice and the pitch data is extracted from the score, so a natural and accurate singing voice can be synthesized.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention is directed to the creation of song speech synthesis data that is input to a speech synthesis circuit that performs rule synthesis of song speech based on speech data consisting of phonological information and prosody information. It is related to the device.

[Prior Art] In general, a speech rule synthesis method is a method that generates speech based on phonetic and linguistic rules by inputting phonological parameters that control phonological information and prosodic parameters that control prosodic information. Although the amount of information to be processed (external input information, internally stored information) is very small, it is a speech synthesis method that can output synthesized speech of a large number of words and sentences.

On the other hand, the text synthesis method is the ultimate speech synthesis method that performs speech synthesis only from character strings. It automatically creates phonological information and prosody information from character strings and performs rule synthesis based on that information. It will be realized. However, although it is currently being actively researched, the quality of synthesized speech is still low.

Here, text synthesis consists of two parts: automatically creating phonological information and prosody information from character strings, storing them as audio data on a storage medium, or outputting them to a transmission line, and the other part: It is conceivable to extract voice data from the system and divide it into two parts, a part that outputs the sound by rule synthesis.In this way, the part that outputs the synthesized speech can be made compact, that is, it can be made smaller, lighter, and lower in cost. (See Japanese Patent Application No. 62-263298).

By the way, when we consider the case of synthesizing singing voices, the ultimate form in this case is that when a musical score is input, singing voices are output. Comparing this with the text synthesis method, it is easy to generate prosodic information because musical scores contain not only character strings (lyrics) but also other information, and the structure is simpler than the text synthesis method. (Patent application 177314/1986)
(see issue).

In this case as well, in the same way as we considered text synthesis, we can automatically create phonological information and prosody information from the musical score, and store it as audio data on a storage medium or output it to a transmission path. It is conceivable to divide the musical score into two parts: a part that extracts audio data from a storage medium or a transmission path, and outputs audio through rule synthesis.6 [Problem to be solved by the invention] However, as mentioned above, it is difficult to If input,
Although the amount of information is larger than a string alone, the height is determined more uniquely than a note, but the length is only given for each note, and it is difficult to determine how much consonant length and vowel length should be used. Since it is not given whether it will be strong or not, it must be determined by rules, and strength is not given as a specific value, so this also needs to be expressed as a specific value according to rules, and at the current technological level, There was a problem that only low-quality audio could be obtained.

The present invention has been made in view of the above points, and its purpose is to obtain voice data for natural singing voices as song voice synthesis data input to a regular synthesis type voice synthesis circuit. Moreover, it is an object of the present invention to provide an inexpensive singing voice synthesis data creation device that can realize voice data for outputting higher quality song voices from a voice synthesis circuit with a simple configuration.

[Means for Solving the Problems] The song speech synthesis data creation device of the present invention provides song speech synthesis data that is input to a speech synthesis circuit that performs rule synthesis of song speech based on speech data consisting of phonological information and prosody information. In the data creation device, a musical score and a singing voice are input, and a musical score information extraction means is provided for extracting character string data as phonological information and height data as prosody information from the input musical score. A song voice information extraction means is provided for extracting strength data and length data, which are prosody information, from the song voice, and character string data, height data, strength data, and length data extracted by both extraction means are provided. It is designed to be output as audio data.

[Function] The present invention is configured as described above, and in a device for creating song speech synthesis data that is input to a speech synthesis circuit that performs rule synthesis of song speech, a character string that is phonetic information is extracted from an input musical score. Provided with musical score information extraction means for extracting data and height data as prosody information, and song voice information extraction means for extracting strength data and length data as prosody information from the input song voice, The character string data, height data, strength data, and length data extracted by both extraction means are output as audio data, and the strength data and length data are extracted based on the actual singing voice. , so that natural singing voice data can be obtained as song voice synthesis data that is input to the regular synthesis type voice synthesis circuit. In addition, since the height data is extracted from the musical score data, even if the pitch of the input singing voice deviates from the musical score, it is possible to synthesize a singing voice with an accurate pitch, and it does not require complicated processing. Since there is no need for pitch extraction processing, it is possible to create voice data for obtaining higher quality singing voices with a simple configuration, and it is now possible to provide an inexpensive device for creating singing voice synthesis data. .

[Embodiment] FIG. 1 shows an embodiment of the present invention, in which song speech synthesis data input to a speech synthesis circuit X that performs rule synthesis of song speech based on speech data consisting of phonological information and prosody information The creation device Y is provided with a musical score information extraction means for inputting a musical score and a singing voice, and extracting character string data as phonological information and height data as prosody information from the input musical score. A song voice information extraction means for extracting strength data and length data, which are prosodic information, from voice is provided, and the character string data, height data, strength data, and length data extracted by both extraction means are extracted from the voice. It is designed to be output as data. Here, the musical score information extracting means is formed by the musical score input section 1 consisting of a manual key device or an optical musical score reading device that inputs the character string data and height data of the musical score, and the phonological information extracting means is formed by a musical score input unit 1 consisting of an optical musical score reading device. a song voice input section 2, a feature extraction section 3 that extracts features of the input song voice, a segmentation processing section 4 that divides the feature-extracted song voice into phonemes, and a strength of each divided phoneme. It is formed by a prosodic information extraction section 5 that extracts data and length data. Further, each extracted data is encoded by the encoding section 6 and outputted as voice data from the voice synthesis circuit X.

Now, in this embodiment, the musical score and singing voice are input, and the voice data (phonological information and prosody information) is output to the voice synthesis circuit X of the regular synthesis method, and the character string is To extract (phoneme sequence) data and height (pitch) data (see Figure 2), there are two ways to do this: a human reading the music score, converting it into data, and inputting it from a keyboard, or automatically using an optical music score reader. Another possible method is to extract data from musical scores.

On the other hand, intensity data and length data are extracted from the input singing voice (see Figure 2). For this purpose, it is necessary to perform segmentation processing to divide the singing voice into each phoneme, and segmentation can be easily performed by using information on the phoneme sequence input from the musical score (Proceedings of the Acoustical Society of Japan). (March 1988) 2-4-1r Automatic creation of phoneme dictionary using top-down phoneme segmentation”)
.

Based on this segmentation result, strength data of each mora and length data of each phoneme are extracted. The data obtained in this manner is encoded by the encoding unit 6, stored in a storage medium, or output to a transmission path, and used as voice data for the voice synthesis circuit X. As described above, data for natural singing voices can be obtained as singing voice synthesis data input to the regular synthesis type voice synthesis circuit, and pitch extraction processing, which requires two complex processes, is not required. , the structure of the device becomes simple and inexpensive.

[Effects of the Invention] The present invention is configured as described above, and in a device for creating singing speech synthesis data that is input to a speech synthesis circuit that performs regular synthesis of singing speech, characters that are phonetic information are extracted from an input musical score. Musical score information extraction means is provided for extracting column data and height data that is prosody information, and song voice information extraction means is provided that extracts strength data and length data that are prosody information from the input song voice. , the character string data, height data, strength data, and length data extracted by both extraction means are output as audio data, and the strength data and length data are extracted based on the actual singing voice. Since the data is extracted, there is an effect that voice data for natural song voices can be obtained as song voice synthesis data that is input to a rule synthesis type voice synthesis circuit. In addition, since the height data is extracted from the musical score data, even if the pitch of the input singing voice deviates from the musical score, it is possible to synthesize a singing voice with an accurate pitch, and it does not require complicated processing. Since pitch extraction processing is not necessary, it is possible to realize voice data for obtaining higher quality singing voices with a simple configuration, and it is possible to provide an inexpensive singing voice synthesis data creation device.

[Brief explanation of the drawing]

FIG. 1 is a block circuit diagram of an embodiment of the present invention, and FIG. 2 is an explanatory diagram of the same operation. X is a speech synthesis circuit, and Y is a generation device for song speech synthesis data.

Claims (1)

[Claims] (1) In a device for creating song speech synthesis data that is input to a speech synthesis circuit that performs regular synthesis of song speech based on speech data consisting of phonological information and prosody information, musical score and singing speech are input. A musical score information extraction means is provided for extracting character string data as phonological information and height data as prosody information from the musical score, and strength data and length data as prosody information are extracted from the input singing voice. Song voice synthesis data characterized in that a song voice information extraction means is provided, and character string data, height data, strength data and length data extracted by both extraction means are output as voice data. Creation device.