JPH06202681A - Speech restoration device - Google Patents

Abstract

PURPOSE:To provide the speech restoration device which can easily recognize the contents of sound data at the time of the reproduction of a speech in synchronism with a moving picture, reproduction processing by a computer which can process data at only a different frequency from the original sound data, or high-speed reproduction for recognizing the outline of the contents of the sound data. CONSTITUTION:The sound data on a speech, etc., generated by data generation processing such as data sampling are divided into plural data units according to the contents (step 101), the divided data units are given priority levels respectively (step 102), and the priority levels and restrictions at the time of reproduction are evaluated (step 103). Then the data units are reproduced according to the sound volume and frequency corresponding to the priority levels.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice restoration device for restoring acoustic data such as voice obtained by means such as data sampling.

[0002]

2. Description of the Related Art Generally, when acoustic data such as voice is handled by a computer, the acoustic data is represented by a list of sampling data according to a certain frequency. The sound data sequentially expressed by the list of the sampling data can reproduce the same sound as the original data by restoring the data at the same frequency as the original data.

With respect to the acoustic data expressed as described above, (1) sound is reproduced in synchronization with a moving image, (2)
In the case of performing operations such as reproduction processing with a computer that can process only at a frequency different from the original sound data, (3) high speed reproduction for grasping the outline of the content of the sound data, conventionally, the following is performed. It was processing.

That is, (a) In the above (1), when the reproduction cannot be performed within the required time, the reproduction is stopped halfway. A method of changing the frequency, and (c) in the above (3), the frequency at the time of reproduction is mechanically increased to perform high-speed processing is adopted.

[0005]

However, each of the above-mentioned conventional methods (a) to (c) has the following problems.

That is, in the above method (a), since the reproduction is skipped, the acoustic data becomes unnatural,
In the method (b), the sound quality deteriorates when the difference between the original frequency and the reproduced frequency is large. Furthermore, in method (C),
If the difference between the reproduction time of the original and the reproduction time is large, the speed becomes too fast, and it becomes difficult to grasp the contents of the audio data such as voice.

In view of the above-mentioned conventional problems, the present invention reproduces sound in synchronization with moving images, or reproduces sound data by a computer capable of processing only at a frequency different from the original sound data. Another object of the present invention is to provide a voice restoration device that can easily grasp the content of the acoustic data when performing an operation such as high-speed reproduction for grasping the outline of the content of the acoustic data.

[0008]

In order to achieve the above-mentioned object, the present invention inputs acoustic data generated by a data generation process including data sampling, and outputs the acoustic data into a plurality of data units according to the contents. Dividing means for dividing, priority means for giving a priority to each of the divided data units, evaluating means for evaluating the priority and constraints during reproduction, and volume according to the priority. And a reproduction processing means for performing the reproduction processing of the data unit with the frequency.

[0009]

According to the present invention having the above-mentioned configuration, the acoustic data such as voice generated by the data generation processing such as data sampling is divided into a plurality of data units according to the contents, and each of the divided data units is divided. The priority is assigned to the data, the priority and the constraint at the time of reproduction are evaluated, and the reproduction processing of the data unit is performed with the volume and the frequency according to the priority. Is easily grasped.

[0010]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing a schematic configuration of an embodiment of an information processing system to which a voice restoration device according to the present invention is applied.

This information processing system is provided with a central processing unit 201 for controlling the operation of the entire apparatus, a voice data 202 stored in a storage unit 204, and priority of each unit of a sentence represented by the voice data 202. Priority data (structure data) 203 expressing a degree, a voice output device 205 such as a speaker, a voice input device 206 such as a microphone, and an input device 207 such as a keyboard.
And a central processing unit 201, a storage device 204, a voice output device 205, a voice input device 206 and an input device 207.
Connected to 08, the main part is composed.

The general flow of the audio reproduction processing of this embodiment is as follows.
It is configured as shown in FIG. The control program of the flowchart of FIG. 2 is stored in the storage device 204, and the processing is executed by the central processing unit 201.

That is, a step 101 for extracting one unit of acoustic data, a step 102 for evaluating the priority of one unit of the extracted acoustic data, and a speed of the acoustic data during reproduction according to the evaluation result of the step 102. And step 103 for determining the size.

Next, the operation of the information processing system configured as described above will be described with reference to FIGS. 3 and 4.

As an example, as shown by 301 in FIG. 3, the voice data "Taro-san tried to go to school with Hanako-san, but because the weather was fine, we went out to play" was restored. Will be described.

[0017] The voice data "Taro-san tried to go to school with Hanako-san, but the weather was fine, so we went out to play with them." It is assumed that three levels of priority are given, separated by a collection of clauses. Note that this priority may be performed by the person who made the input, or the system may analyze the meaning of the sentence and automatically perform it. However, in this description, the person who inputs the voice data is the unit of the sentence. The description will be made assuming that the division is performed and the priority order thereof is specified.

At the time of inputting voice data, the voice registrant inputs the voice data from the voice input device 206 shown in FIG. 1 by voice by activating the data input program of the central processing unit 201. The input voice is then processed by the central processing unit 20.
It is recognized by the voice recognition program that operates in 1, and is converted into the form of a sentence containing kanji and kana. Although this conversion into a kanji / kana mixed sentence is not essential in the present invention, it will be described for the sake of explanation in order to give a better user interface. At this point, if the voice recognition program makes an erroneous recognition, or if proper kana conversion is not performed, it is possible to edit the kanji / kana mixed sentence by the editing function of the ordinary computer system.

Next, the prioritization program is operated to divide the sentence into unit parts and prioritize the Kanji / Kana mixed sentence thus obtained. In a preferred embodiment of the present invention, a sentence is divided into meaningful units each consisting of a group of phrases, and a priority is added to the unit. The priority of the priority program is
For example, it is given by first entering a mode of giving a first priority, and in this mode, by sequentially clicking a clause group having a first priority with a mouse.

To explain this point as an example, in the mode of giving the first priority, first, when the mouse is placed on the phrase "Taro san" and the click is performed, that part is highlighted. , Indicates that it has been selected. Then click the part "I went to play with two people",
This part is also highlighted, indicating that it has been given the first priority. The end of the designation of the first priority is
For example, a "designation end" button (not shown) may be clicked with the mouse.

Next, the mode for designating the second priority is entered, and similarly, the portion having the second priority is designated. When specifying the phrase, if the system recognizes the phrase in advance, you can give priority to any part of the phrase by clicking, but in the case of a system that does not recognize the phrase, it is usually done as usual. The priority range may be specified by a combination of mouse click and drag.

Sentences whose priorities have been designated as described above have the data structure shown in FIG.
It is stored in.

That is, the audio data is divided into unit parts and stored in the storage device 204, and the priority data for each unit part is stored in the storage device 204 while holding two pointers. The two pointers held by the priority data are a pointer to audio data corresponding to the priority data and a pointer to priority data for the next unit of audio data.

As the pointer to the priority data for the last voice data unit, it is assumed that a special pointer indicating the last data is stored. In the actual storage, first store the priority of the unit of the beginning part of the audio data and the unit of the beginning part of the actual audio data in the file, then store the priority of the second unit and the actual audio data. The priority and the actual data are alternately stored, such as storing. Here, each pointer is converted into a numerical value indicating how many data in the file should be skipped and stored. A special pointer that represents the last data is represented by the number 0.

Next, the case of reproducing a sentence having the voice data and the structure data thus stored will be described.

In the present description, a case where the user designates the data 301 in FIG. 3 to be reproduced up to priority 1 will be described as an example.

When activating the reproduction program, the user specifies the audio data to be reproduced and the priority of reproduction. In the reproduction process of the audio data, the strength and speed of the volume at the time of reproducing the audio data may be determined according to the priority. In this description, a method of determining whether to reproduce or skip without reproducing according to the priority will be described.

When the reproduction program is started, the reproduction program opens (1) the file in which the designated audio data is stored, extracts the priority data stored at the beginning, and (2) gives priority. The variable, the corresponding voice data, and the pointer to the next priority data are stored in variables. (3) The read priority is compared with the priority designated by the user. As a result of this comparison, if the read priority is the same as or smaller than the designated priority, the reproduction processing of the audio data corresponding to the read priority is performed, and if not, the reproduction processing is not performed. In this example, since the read priority 1 is the same as the priority 1 designated by the user, the reproduction process is performed.

(4) Looking at the stored pointer to the next priority data, if this is the pointer representing the last data, the processing is terminated, and if not, the next priority data Again, the same processes as the above processes (2) to (4) are repeated. In this example, since this is not the last data, the next priority data is fetched based on the pointer to the next priority data, and the processes from (2) above are repeated.

The reproduction process in the above description means that the audio data is extracted and the audio output device 205 of FIG. 1 is used to reproduce the audio. Further, since each pointer stores the number of data to be skipped in the storage device 204, in order to retrieve the data indicated by the pointer, the necessary number of data is skipped before the data in the storage device 204 is skipped. Should be read. As a result of the above processing, the audio data reproduced in this example is 303-in FIG.
It becomes as shown in 1. It should be noted that the same audio data as described above for the reproduction program and the result of the reproduction processing when priority 2 is given as the input of the user are shown in 303-2 of FIG. 3, and the same audio data as the above for the reproduction program. And priority 3
303-3 in FIG. 3 shows the results of the reproduction process when is given as the input by the user.

[0031]

As described above, according to the present invention, the acoustic data generated by the data generation process including the data sampling is input, and the acoustic data is divided into a plurality of data units according to the contents. Dividing means,
Priority ordering means for giving priority to each of the divided data units, evaluation means for evaluating the priority order and constraints at the time of reproduction, and the data unit with the volume and frequency according to the priority order. Since the audio data is reproduced in synchronization with the moving image, or the reproduction processing is performed by a computer capable of processing only at a frequency different from that of the original audio data. Alternatively, when an operation such as high-speed reproduction for grasping the outline of the contents of the acoustic data is performed, the contents of the acoustic data such as voice can be grasped more easily than the conventional method.

[Brief description of drawings]

FIG. 1 is a block diagram showing a schematic configuration of an embodiment of an information processing system to which a voice restoration device according to the present invention is applied.

FIG. 2 is a flowchart showing a schematic flow of a voice reproduction process of the embodiment.

FIG. 3 is a diagram for explaining a specific example of the audio reproduction process of the above-described embodiment.

FIG. 4 is a diagram showing a data structure of a sentence having a designated priority stored in a storage device.

[Explanation of symbols]

 201 Central Processing Unit 202 Voice Data 203 Structure Data 204 Storage Device 205 Voice Output Device 206 Voice Input Device 207 Input Device

Claims (1)

[Claims] 1. A dividing unit for inputting acoustic data generated by a data generation process including data sampling and dividing the acoustic data into a plurality of data units according to the contents, and to each of the divided data units. Priority means for giving priority to each other, evaluation means for evaluating the priority and constraints at the time of reproduction, and reproduction processing means for performing reproduction processing of the data unit with a volume and frequency according to the priority. A voice restoration device characterized by comprising.