JPH04304500A - Voice information reproduction control system - Google Patents

Abstract

PURPOSE:To automatically retrieve and classify voice information by analyzing the voice information previously and storing information which is required for reproduction control corresponding to source voice information one to one. CONSTITUTION:A listener who desires to retrieve the voice information registers a necessary key word in an audiotex center 100. A sound recording service offerer for information records the voice information in a voice information storage part 120 together with information on a classification code, etc., required for the control of a reproduction control part 140. The stored information is stored in a reproduction control information storage part 130. Then when origination from the listener is connected, the reproduction control part 140 takes the necessary voice information out according to a received operation code and reproduces it, and the jumping or tracing-back reproduction of a necessary section in the voice information is performed. Thus, voice recognition technology is introduced to enable speedy retrieving operation, evade the word head cutting, etc., at the time of the jumping and tracing-back reproduction, and automatically retrieve and classify the voice information, thereby improving the operability of the audiotex service.

Description

[Detailed description of the invention]

0001

[Industrial Application Field] The present invention provides information that is stored as audio information to a person who wishes to listen to the information (hereinafter referred to as a listener).
This is a technology related to information retrieval for an audiotex system that plays based on the operation of a user.

0002

2. Description of the Related Art FIG. 6 is a diagram illustrating a conventional technology, in which 10 is an audiotex center, 20 is a network
30 is a terminal, 11 is an audio information storage section, and 12 is a playback control section.

[0003] Conventionally, when a listener wants to listen to stored audio information, a terminal (usually a PB telephone) 30
A call is made to the Audiotex center 10 for connection. The connected audiotex center 10 extracts guidance necessary for operation from the audio information storage section 11 under the control of the playback control section 12 and sends it out.
Receives the operation instruction code signal from the listener. DTMF (two-frequency mixed signal) usually emitted from PB telephones
receive. If the listener does not know the content of the stored audio information, the playback control unit 12 provides guidance on the information regarding the stored audio information and the corresponding instruction code, and the listener follows this guidance to input the desired information. Listen.

0004

[Problems to be Solved by the Invention] However, as the amount of information increases, it becomes necessary to listen to a large amount of guidance before listening to the desired information. For this reason, the information content is divided into large items, medium items, and small items according to the field, and is classified into tree-shaped items to control the information so that the desired information can be reached as quickly as possible. After specifying
The necessary information cannot be heard unless all of the information contained in the sub-items is listened to in order, which again requires a great deal of time.

SUMMARY OF THE INVENTION An object of the present invention is to provide an audio information reproduction control system that allows desired information to be quickly extracted and listened to from a large amount of audio information.

[0006]

[Means for Solving the Problems] The invention according to claim 1 of the audio information reproduction control method of the present invention provides reproduction control information that analyzes audio information, extracts and outputs reproduction control information necessary at the time of reproduction. an analysis extraction section, a playback control information storage section that stores a plurality of pieces of playback control information, an audio information storage section that stores original audio information corresponding to the playback control information, and an operation code that receives the operation code and performs the necessary playback operations. It is equipped with a playback control section for controlling.

[0007] Further, in the invention as set forth in claim 2, the playback control information analysis and extraction section divides the audio information into fixed frames, determines the presence/absence of each frame, and divides the audio information into fixed frames. When consecutive frames are silent, the position of any frame in the consecutive frames is registered in the playback control information as a playback break position, and the playback is performed by skipping or going backwards over a certain audio section. When the operation control unit receives an operation code for performing the above operation, the operation control unit controls the playback to start from the playback break position adjacent to the skip destination or backward playback position.

[0008]Furthermore, in the invention according to claim 3, the playback control information analysis and extraction section includes a speech recognition section capable of recognizing specific words, registers words to be recognized in advance, and uses the accumulated speech information. Recognize the registered word from among them, register a code that specifies audio information that includes the registered word as playback control information, and include the registered word when a specific operation code is received. It reproduces only audio information.

[0009]

[Operation] In the invention as set forth in claim 1 of the present invention, a playback control information analysis and extraction section is added that analyzes and extracts the accumulated audio information and outputs the information necessary for playback control. By storing such information in correspondence with the original audio information, playback can be started quickly in response to various operation requests necessary for searching.

[0010] Various operations can be considered as necessary for searching, but typical examples include skipping/backward playback within one piece of audio information, and a search function that specifies only the necessary one among multiple audio information. It will be done. Regarding the former, one possible method is to skip a certain interval (5 seconds or 10 seconds) or go back in time for a single instruction, but if you simply control by time, the playback start point will be in the middle of the word. It becomes difficult to understand. In order to avoid this and at the same time cover the delay caused by processing online after receiving the operation code, the invention as claimed in claim 2 detects the break between words in advance and uses it as control information. By controlling the playback based on this, it is possible to perform easy-to-listen and quick skip or backward playback. Regarding the latter, the invention according to claim 3 includes a voice recognition unit that can register keywords related to necessary information in advance,
To quickly listen to necessary information by providing a listener with a function of sequentially reproducing only audio information including the keyword.

[0011]

DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 is a diagram illustrating an embodiment of the present invention, in which 100 is an audiotex center, 20 is a network, 30 is a terminal, 110 is a reproduction control information analysis and extraction unit, and 120 is a 130 is a reproduction control information storage unit, 140 is a reproduction control unit, 111 is a voice/silence extraction unit,
112 is a speech recognition section. In this embodiment, a user (a person who records and plays information) makes a call connection to the audiotex center 100 from the terminal 30, stores information according to the guidance, and listeners also make a call in the same way. This shows a case where the user connects, inputs the operation code according to the guidance, and listens to the desired information.

[0012] A listener who wishes to search for audio information calls and connects to the audiotex center 100 in advance.
The necessary keywords are registered in the voice recognition section 112 according to the guidance instructions. On the other hand, when information is stored, that is, when an information recording service is provided, the necessary guidance is created by the control of the playback control unit 140, and the information provider (among the users) The person using the recording service inputs the necessary information such as the classification code and records the audio information in the audio information storage section 1.
Record at 20. The audio information once stored in the audio information storage section 120 is sent to the reproduction control information analysis and extraction section 110, where the reproduction control information is extracted.
is accumulated in Thereafter, when there is an outgoing connection from the listener, the playback control unit 140 extracts and plays back the necessary audio information based on the received operation code, or skips over or plays back the necessary sections of the audio information. conduct.

The internal control system will be explained in detail below. Normally, when controlling the transfer of audio files stored on a storage medium such as a disk, the audio file directory is divided into two layers, and address information is assigned to each area that stores a certain length of audio information, as shown in Figure 2. Manage as a cell. When trying to realize a skip/backward function with such a configuration (claim 2), when a skip/backward instruction is received, the address in the cell being played is read, the necessary shift value is calculated, and the corresponding cell is It is necessary to search for the cell, read out the audio data in the cell, and perform audio information processing. Specifically, the read audio information is divided into frames of appropriate length, and the power of each frame is calculated to determine whether there is a sound or not. Then, by detecting a certain number of consecutive silent frames, the audio data address corresponding to the frame is obtained. This control flow is shown in FIG. The voice information processing portion surrounded by the dotted line in the figure requires a large amount of processing compared to other portions, and if this is performed online on-demand, the response time will be long. If word spotting processing is attempted (claim 3), the amount of processing will be several orders of magnitude or more, and since it must be performed for all audio files, the response time will further increase.

On the other hand, if a method is adopted in which control information is extracted in advance as described above, the response time can be reduced by an order of magnitude. A specific example of reproduction control information is shown in FIG. By incorporating the delimiter cell address and the relative address value from the top address into the basic audio file management information shown in FIG. 2, the flow in FIG. 3 is simplified as shown in FIG. 5. Furthermore, if keywords are extracted for each file and the information is coded and stored, searching for audio files that have the corresponding keyword information will be easier than voice recognition processing. is extremely simplified.

In this embodiment, voice recognition is used to extract voice information that includes a desired keyword from a plurality of pieces of voice information. It can also be used for field classification. If this is possible, the classification code will be used when information providers accumulate information.
This saves the trouble of instructing the user to follow the guidance, and improves operability. Further, the details of the method of registering words necessary for speech recognition are not described because they are not directly related to the present invention, but may vary depending on the method and performance of the recognition device employed. For example, there are methods in which you enter a character-compatible code to specify the necessary keywords, methods in which you specify the necessary words from a predetermined set of words,
Alternatively, a method of giving instructions using the registrant's own voice may also be considered. In addition, as voice recognition performance improves and it becomes possible to understand the meaning of the entire voice information, various improvements can be made, such as automating the classification of information and editing a summary of the content and providing it to the listener. Since quality is the most important factor in terms of service performance, the method of analyzing and extracting playback control information in advance through off-line processing as shown in the present invention is generally effective.

[0016]

[Effects of the Invention] As explained above, in the present invention, regarding the search and playback of voice information, the accumulated voice information is analyzed in advance, and information necessary for playback control is extracted and stored, thereby allowing quick search. At the same time, we focused on the characteristics of the voice as playback control information, making it easier to listen to by avoiding truncations at the beginning of words even when skipping or going backwards, and by introducing voice recognition technology, we were able to improve voice control, which was previously impossible. It enables automatic search and classification of information, making it possible to greatly improve the operability of Audiotex services.

Furthermore, the playback control information analysis and extraction unit of the present invention detects word breaks in advance and stores them as control information, and controls playback based on this information, so that easy-to-listen and quick skipping or backward playback is possible.

Furthermore, the reproduction control information analysis and extraction section of the present invention registers keywords related to necessary information in advance, so that necessary information can be quickly heard.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing one embodiment of the present invention.

FIG. 2 is a diagram illustrating the operation of FIG. 1;

FIG. 3 is a flowchart showing the manufacturing flow of the embodiment of FIG. 1.
It is.

FIG. 4 is a diagram showing a specific example of reproduction control information used in the present invention.

FIG. 5 is a flowchart showing a control flow of another embodiment of the present invention.
It is a chart.

FIG. 6 is a block diagram showing a conventional audio information reproduction control method.

[Explanation of symbols]

20 Network 30 Terminal 100 Audiotex Center 110 Reproduction control information analysis and extraction section 111 Sound/silence extraction part 112 Voice recognition section 120 Audio information storage section 130 Playback control information storage unit 140 Playback control section

Claims (3)

[Claims] 1. A playback control information analysis and extraction unit that analyzes audio information and extracts and outputs playback control information necessary for playback;
a reproduction control information storage unit that stores a plurality of pieces of reproduction control information;
It is equipped with an audio information storage section that stores original audio information corresponding to playback control information, and a playback control section that receives operation codes and controls necessary playback operations, and when audio information is added or changed, The added or changed audio information is analyzed by the playback control information analysis and extraction unit to extract the playback control information, and the playback control unit adds or changes the corresponding stored playback control information to generate the audio. When information is deleted, the corresponding stored playback control information is deleted, and the playback of the audio information is controlled using the corresponding playback control information based on the received operation code. Audio information playback control method. [Claim 2] The playback control information analysis and extraction section divides the audio information into fixed frames, determines the presence/absence of each frame, and determines whether a certain number of consecutive frames are silent or not. In this case, the position of any frame of the continuous frames is registered in the playback control information as a playback break position, and the operation code for skipping over a certain audio section or going back and playing the playback is controlled. 2. The audio information playback control system according to claim 1, wherein when the audio information is received by the audio information playback control system, the audio information playback control method is controlled to start playback from a playback break position adjacent to a playback position of a jump destination or a backward playback destination. 3. The playback control information analysis and extraction unit includes a voice recognition unit capable of recognizing specific words, registers words to be recognized in advance, and recognizes the registered words from the accumulated voice information. A code specifying audio information that includes the registered word is registered as playback control information, and when a specific operation code is received, only audio information that includes the registered word is played back. 2. The audio information reproduction control system according to claim 1.