JPH07210194A - Device for outputting sound - Google Patents

Abstract

PURPOSE:To accelerate a process time until an sound output of recorded and reproduced sound and to perform smoothly the timing control of the sound output between the sound recorded and reproduced and a rule synthesis sound. CONSTITUTION:A sound output device is provided with an output response part generating a required output response code from output information from a position senser and a path search part, and is connected to a device 101 receiving required information and an output device 102 such as a speaker, etc., and the main body is constituted of a control part 103 controlling the whole device, a large scale sound recording file 104, a high speed access memory 105 beforehand loading the sound data of important words from the large scale sound recording file 104, a high speed memory control part 106, a sound recording/reproducing part 107, a rule synthesis part 108, a phoneme element memory 109 and a connection output part 110 outputting the sound.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice output device by voice synthesis, which is used for various voice guidances and the like.

[0002]

2. Description of the Related Art When a character string is input from an input device, if the character string is a main character string, the voice data corresponding to the character string is read from a memory storing the voice data, and the voice data is also stored. And PCM are used for voice synthesis. When a character string whose voice data is not stored in the memory in advance is input, linguistic analysis is performed to obtain phonological information such as reading of the character string and prosodic information such as pauses and accents between words, and There is a voice output device that performs rule synthesis based on prosody information and outputs a voice corresponding to an input character string.

[0003]

However, although the voice data is searched for the information to be voice output,
There is a problem in that there is a difference between the time required to output the voice of the regular synthesized voice and the time required to search and output the voice of the recording / playback sound, which causes a problem in the timing control of the voice output.

Further, there is a difference in pitch information between the voice generated by rule synthesis and the recorded and reproduced sound, that is, the pitch frequency, and when the voices generated by both formulas are connected and output as a sentence voice, There is a problem that the quality of voice deteriorates.

[0005]

SUMMARY OF THE INVENTION The first object of the present invention is to quickly output a voice in response to a voice output command.
As a concrete means, a high-speed access memory is prepared in addition to large-scale audio data, and a means for transferring and holding important audio data that is highly likely to be used for audio output in the high-speed access memory is provided. Some voices are output by recording and playback, and other voices are output by rule synthesis.

Further, the second purpose is to output a high quality voice, and a means for adjusting pitch, power, phoneme duration etc. is provided to output a natural sentence voice.

[0007]

In the above voice output device, when the information such as the position on the map is input, the high speed access memory is searched for the voice data corresponding to the input information, and if the voice data is present, the recording is performed. The reproduction unit reproduces the voice data, and when there is no voice data in the high-speed access memory, the rule synthesis unit reads out the phoneme unit data from the phoneme unit memory and performs rule composition. Further, the connection output unit reads out and connects each of the divided sounds generated as described above, and outputs the sound from the output device.

[0008]

DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 is a block diagram of a system of an audio output apparatus for recording / playback / rule synthesis combined use for an in-vehicle navigation system showing an embodiment of the present invention. This voice output device has an output response unit that generates a necessary output response code from the output information of the position sensor and the route search unit in the vehicle-mounted navigation device, and the voice navigation is composed of the character code output from the output response unit. A device 101 for receiving necessary information, a main body, and an output device 102 such as a speaker for outputting voice are connected.

The main body includes a control unit 103 for controlling the entire apparatus, a large-scale recording file 104 for storing sentence voices including a place name and an intersection name necessary for voice navigation in a plurality of divisions, and a large-scale recording file 104. A high-speed access memory 105 that pre-loads voice data of important words from the large-scale recording file 104, a high-speed memory control unit 106 that transfers voice data from the large-scale recording file to the high-speed access memory at any time, and a high-speed access memory A recording / reproducing unit 107 for reproducing a voice, a rule synthesizing unit 108 for generating and outputting a voice that does not exist in the high-speed access memory, a phoneme unit memory 109 for storing voice unit data necessary for generating a ruled synthesized sound, It is composed of a connection output unit 110 that reads out segmented voices, connects them, and outputs voices.

According to the flow chart shown in FIG.
The outline of the operation of the recording / playback / rule synthesis combined type voice output device for the vehicle-mounted navigation system shown in FIG.

When a voice output command is issued in the vehicle-mounted navigation device, a position sensor or the like in the vehicle-mounted navigation device transmits a character code such as position information necessary for performing navigation by voice to the character code receiving device 101. . A character code is output from the character code receiving device 101 to the control unit 103, and it is searched whether a voice corresponding to the character code is stored in the high speed access memory 105.

If the voice corresponding to the character code is stored in the high speed access memory 105, the voice is reproduced by the recording / reproducing unit 107. On the other hand, if the voice corresponding to the output character code is not stored in the high-speed access memory 105, first, the control unit 103 converts the character code into phoneme / accent information, and the rule synthesizing unit 108 based on the information. The phoneme unit data that is a phoneme is read from the phoneme unit memory 109 and rule synthesis is performed. Character information / accent information corresponding to all character codes is stored in the high-speed access memory 105. Further, the connection output unit 110 reads out and connects the respective divided sounds generated as described above, and outputs the sounds from the output device.

Next, control for transferring audio data from the large-scale recording file 104 to the high-speed access memory 105 will be described. FIG. 3 shows an example of the data structure of a large-scale recording file. Since the text of the voice guidance in the vehicle-mounted navigation includes a lot of geographical information such as intersection names and place names, such voice data is stored together with the character code indicating the position on the map. Furthermore, regarding the audio data in this large-scale recording file, the area is divided so that several audio data are included based on the character code indicating the position on the map, etc., and the audio data is stored for each block. Keep it. Transfer of audio data from a large-scale recording file to the high-speed access memory is performed for each block.

Next, control for selecting important voice data to be transferred from the large-scale recording file 104 to the high speed access memory 105 will be described. FIG. 4 shows the correspondence between a map used for in-vehicle navigation and blocks of audio data in a large-scale recording file. Each area separated by a dotted line on the map corresponds to a block in which audio data is divided and stored in a large-scale recording file. The selection of blocks to be transferred is determined based on conditions such as the current location, the destination, and the direction of the road on which the vehicle is running. In principle, the audio data of eight blocks surrounding the block including the current location are transferred to the high speed access memory. Taking the case of FIG. 4 as an example, an example of selecting blocks to be transferred will be described with reference to the flowchart of FIG.
The block where the current position is is C4. That is, the audio data of the eight blocks surrounding C4 are transferred to the high-speed access memory. Next, based on conditions such as the direction of the road and the destination, the block that can be moved from the current position is C3.
It can be limited to D3. For example, when the vehicle travels according to the recommended route of the navigation system, five blocks C2, D2, E2, E3 and E4 are selected as blocks to be transferred.

When the vehicle goes straight on the road on which the vehicle is running, three blocks B2, C2 and D2 are transferred.

In parallel with this processing, unnecessary voice data in the high speed access memory is also selected for each block. Blocks of unnecessary audio data are judged based on conditions such as the current location, the destination, and the direction of the road on which the vehicle is running.

Next, an embodiment of the data structure in the high speed access memory is shown in FIG. In the high-speed access memory, intersection names and place names corresponding to character codes indicating positions on the map are stored together with phonological accent information and the like. When the voice data is transferred from the large-scale recording file, it is stored in the voice data area of the corresponding position code. In addition, for fixed phrases that are frequently used in in-vehicle navigation systems, such as "Next ... Turn right at the intersection.", The recorded voice is divided into multiple sections and stored in the high-speed access memory beforehand. Store.
Regarding the voice data of such a fixed sentence, the values of pitch, power, and phoneme duration are stored for each segmented voice. FIG. 7 shows an embodiment of a data structure relating to audio data that is frequently stored in the high speed access memory.

Next, control for smoothly connecting the divided voices will be described. When using a standard sentence as the output sentence, smoothing the connection between the segmented voice (hereinafter referred to as the variable part voice) indicated by the broken line in the above example and the segmental voice of the fixed phrase adjacent to the variable part voice is natural. The sentence sound is output. Therefore, the connection output unit adjusts the pitch, power, phoneme duration, etc. of the variable part voice based on the information such as the pitch, power, phoneme duration, etc. of the segmented speech of the adjacent fixed sentence, and performs connection processing. Output audio. FIG. 8 is a diagram showing a processing concept of the connection processing.

When the variable part speech is generated by the rule synthesizing part, the rule synthesizing is carried out beforehand based on the information such as the phoneme accent information and the pitch, power, and phoneme duration of the segmented speech of the adjacent fixed sentence. May be.

[0020]

According to the present invention, since the recording data to be output is stored in the high speed access memory, the processing time until the voice output of the recording / reproducing sound is accelerated, and the recording / reproducing sound and the regular synthesized sound are combined. The audio output timing control can be smoothly performed. Since the voice data for recording and reproduction requires a large storage capacity, the speed of the processing according to the present invention is remarkable. Furthermore, since the pitch, power, and phoneme duration length values are adjusted for the voice data and the voice data is connected, a high-quality text voice is provided.

[Brief description of drawings]

FIG. 1 is a block diagram of a system of an audio output device for recording / playback / rule synthesis combined use for an in-vehicle navigation system showing an embodiment of the present invention.

FIG. 2 is a flow chart of the operation of the recording / playback / rule synthesis combined type voice output device for an in-vehicle navigation system showing an embodiment of the present invention.

FIG. 3 is an explanatory diagram of a data structure of a large-scale recording file for an in-vehicle navigation system showing an embodiment of the present invention.

FIG. 4 is an explanatory diagram showing the correspondence between a vehicle-mounted navigation map and a block of audio data of a large-scale recording file showing an embodiment of the present invention.

FIG. 5 is a flowchart of an embodiment of selecting a block of audio data to be transferred from a large-scale recording file to a high-speed access memory according to the present invention.

FIG. 6 is an explanatory diagram showing an example of a data structure in a vehicle-mounted navigation high-speed access memory according to the present invention.

FIG. 7 is an explanatory diagram showing an embodiment of a data structure relating to frequently-used voice data stored in a vehicle-mounted navigation high-speed access memory according to the present invention.

FIG. 8 is an explanatory diagram showing a processing concept of a connection processing of an audio output device for a vehicle-mounted navigation device showing an embodiment of the present invention.

[Explanation of symbols]

101 ... Character code receiving device, 102 ... Voice output device,
103 ... Control unit, 104 ... Large-scale recording file, 105 ...
High-speed access memory, 106 ... High-speed access memory control unit, 107 ... Recording / playback unit, 108 ... Rule synthesis unit, 109
... phoneme unit memory, 110 ... connection output unit.

Front page continuation (72) Inventor Nobuo Hataoka 1-280 Higashi Koigokubo, Kokubunji, Tokyo Inside Hitachi Central Research Laboratory (72) Inventor Shigeru Kakumoto 1-280 Higashi Koikeku, Kokubunji, Tokyo Hitachi Central Research Co., Ltd. In-house (72) Inventor Tomoko Hatakeyama 1-280 Higashi-Kengikubo, Kokubunji-shi, Tokyo Inside Central Research Laboratory, Hitachi, Ltd.

Claims (8)

[Claims] 1. A large amount of text voice is divided into a plurality of sections in units of phrases, and waveform data storage means for storing waveform data corresponding to the text voice, and section voice waveform data is selected from the waveform data storage means. And a high-speed access memory for storing the waveform data of the segmented voice selected from the waveform data storage unit, a unit for reading out each segmented voice from the high-speed access memory and reproducing and outputting the recorded voice, and a character code string. In the device which outputs the sentence voice by connecting the respective segmented voices generated by the means for performing the regular voice synthesis for generating the voice waveform signal from the sentence voice, the segmented waveform data of the sentence voice to be output exists in the high-speed access memory. Means for determining whether or not the section waveform data does not exist in the high-speed access memory, and a sentence corresponding to the section A voice output device comprising means for performing rule synthesis from a character code string to obtain segmented voice waveform data. 2. The means for transferring the segmented speech waveform data selected from the waveform data storage means to the high speed access memory, and the segmented speech waveform data transferred from the waveform data storage means at the high speed according to claim 1. Means for accumulating in the access memory, and means for storing the number of times of reading of the divided voice waveform data and the storage time in the high speed access memory as the usage frequency information of each divided voice waveform data transferred to the high speed access memory. The means for determining the necessity / non-necessity of the segmented voice waveform data in the high-speed access memory based on the usage frequency information of the respective segmented voice waveform data transferred to the high-speed access memory, and the voice waveform data determined to be unnecessary are And a means for deleting from the high speed access memory. 3. A large amount of text speech is divided into a plurality of sections in units such as clauses, and characteristic parameter storage means for storing characteristic parameters of speech of LPC parameters corresponding to the text speech, and classification by the characteristic parameter storage means. And a high-speed access memory for storing the feature parameters of the segmented voice selected from the feature parameter storage unit, each segmented voice is read from the high-speed access memory, and the recorded voice is reproduced and output. In the device which outputs the sentence voice by connecting the respective segmented voices generated by the means and the means for performing the regular voice synthesis for generating the voice waveform signal from the character code string, the characteristic parameter of the segmented voice of the sentence voice to be output is The means for determining whether or not it exists in the high speed access memory, and A voice output device comprising means for obtaining a feature parameter of a segmented voice by performing rule synthesis from a character code string corresponding to the segment when it does not exist in the high speed access memory. 4. The means for transferring the feature parameter of the segmented voice selected from the voice feature parameter storage means to the high speed access memory, and the feature parameter of the segmented voice transferred from the waveform data storage means according to claim 3. Means for accumulating in the high-speed access memory, and the frequency of reading the characteristic parameters of the segmental voice and the storage time in the high-speed access memory as usage frequency information of the characteristic parameters of each segmental voice transferred to the high-speed access memory. Means for storing, means for judging the necessity / non-necessity of the characteristic parameter of the divided voice in the high-speed access memory based on the use frequency information of the characteristic parameter of each divided voice transferred to the high-speed access memory And a means for deleting from the high speed access memory. An audio output device, wherein the door. 5. The voice segment according to claim 1, 2, 3 or 4, which is frequently used, is a pitch value together with voice data.
Means for storing prosody information such as power value and phoneme duration in the high-speed access memory;
A voice output device comprising means for adjusting the values of pitch, power, and phoneme time length, and connecting the voice generated by rule synthesis or recording / playback to output voice. 6. The vehicle-mounted navigation device according to claim 5, which has a means for detecting a vehicle position, a means for searching for a route, and a means for guiding a route, and uses voice as information presenting means such as route information. 7. An area on a map is divided into several parts and defined as a block, and a position on the map where voice data is output is defined as a position code, and a position code in the same block is defined in claim 6. An in-vehicle voice navigation device that stores voice data corresponding to the position code and the block code as one set. 8. The vehicle-mounted voice navigation device according to claim 7, wherein the means for storing in the high-speed access memory is a block unit based on the information of the current position.