KR101904817B1 - Call conversation Speech to Text converting system - Google Patents

Abstract

The present invention relates to a call conversation speech to text converting system. More particularly, the present invention relates to a call conversation speech to text converting system which separates speech from the call conversation of a receiver and divide and store a caller and the receiver when making a call using voice over Internet protocol (VoIP), separates and stores the silent interval of the collected speech packets of the caller and the receiver into segments to improve the accuracy of speech recognition, converts the separated speech packet into text, arranges sentences in time series based on time information applied as the time stamp of each packet and create a deposition when integrating, to facilitate text search with words and sentences, and accumulates repeatedly depositions to obtain big data. The call conversation speech to text converting system includes a server, a packet collecting part, a speech separating part, a text conversing part, and a text integrating part.

Description

{Call conversation Speech to text converting system}

BACKGROUND OF THE INVENTION 1. Field of the Invention [0002] The present invention relates to a voice-to-text conversion transcription recording system. More specifically, voice is separated from the contents of a call, a call, and a recipient by using an Internet telephone (VoIP), and the voice and the voice are separately classified into a sender and a receiver. The speech recognition accuracy can be improved by segmenting and segmenting the sentences, and the separated speech packets can be converted into texts, and when they are integrated, transcriptions can be arranged in time series based on time information applied as time stamps of respective packets, Text conversion transcription recording system capable of facilitating text search with sentences and securing big data at a repetitive transcription scale.

Currently, as the use of the Internet becomes more common with the development of technology, a VoIP system is being actively developed as an Internet telephone service capable of making a telephone call over the web. Such an Internet telephone service uses an Internet communication network instead of a conventional public telephone network (PSTN), and converts an analog voice signal into a digital packet voice signal. The Internet telephony service is a packet-switched network that performs dynamic routing and provides a "best-effort" type service, which makes it difficult to guarantee call quality. However, Internet telephony service is being used in various fields such as call center, public institution, corporation (communication company), and outsourcing company which require a large amount of telephone calls because it is possible to make long distance calls at low prices and create various types of services.

On the other hand, there is an event processing device as an apparatus for interfacing a telephone of a call center and a computer. An example of an event processing device is a computer telephony integration (CTI) system. The CTI system is a system that integrates a computer and a telephone, and controls the exchange using a PC.

The event processing unit efficiently informs the agent of a lot of information related to the customer through the telephone exchange network and the IVR (Interactive Voice Response), such as the customer's telephone number, card number, resident number, service code and other additional key values, It makes it possible to conveniently manage telephone counseling tasks using computers.

For example, if a customer wants to talk to a specific company for consultation, the event processor will show a lot of information about the customer on the agent's monitor, so that smooth and prompt consultation can be done and the agent needs to call We also automatically arrange the data for customers who have a phone call.

The exchanger and the event processing device are connected through a specific link, the event processing device and the database host are connected to the network (LAN), and the agent is connected to the event processing device or the DB host Send and receive progress status information. A device that can record telephone conversations between a customer and an agent, record the contents of the response, and reproduce the recording will be a clear means of evidence in case of future disputes.

However, there is a limit to the storage capacity of the recorded file by recording the contents of the response described above, and it is difficult to search for a specific phrase in the future, and there is a limitation in using the second data through data analysis in the future have.

Prior Art Document: KR Patent Registration No. 1213514 (published on Dec. 18, 2012)

SUMMARY OF THE INVENTION The present invention has been made in order to solve the above-mentioned problems, and it is an object of the present invention to provide an apparatus and method for searching for a text content file, The present invention provides a voice-to-text transcription record generation system capable of voice recognition and high accuracy of text extraction even in a continuous sentence.

According to an aspect of the present invention, there is provided an apparatus for generating a voice-to-text conversion transcription record according to the present invention. The system includes an agent for providing a consultation service using an Internet telephone (VoIP) A server for storing, managing, and providing call details of the server; A packet collecting part for collecting voice packets, the packet collecting part collecting voice packets of a consultant separately from voice packets of a service requester; A voice separator for storing voice data obtained by dividing voice packets collected by the packet collecting unit into segments; A text conversion unit for converting the speech data of the segment unit separated by the speech separation unit into text; And a text integration unit for integrating the text converted by the text conversion unit and storing the integrated text in the DB.

In addition, the voice separating unit may include separating voice packets collected by the packet collecting unit on a segment-by-segment basis, and separating voice packets based on a silence interval.

In addition, the speech separator may include a segment time stamp to which a time point of data is displayed for each of the separated segments.

According to the present invention, voice is separated from the contents of a call, a call, and a receiver by using an Internet telephone (VoIP), and the voice and the voice are separately classified into a sender and a receiver. And the speech recognition accuracy can be improved by separately storing the speech data in segment units.

In addition, according to the present invention, when the separated speech packets are converted into texts and integrated, the transcriptions can be arranged in a time-lapse manner on the basis of the time information applied as the time stamp of each packet, And it is possible to secure big data at a repetitive transcript scale.

1 is a conceptual diagram of a voice-to-text conversion transcription recording system according to a preferred embodiment of the present invention;
FIG. 2 illustrates a header structure of an RTP packet in a voice-to-text conversion transcription recording system according to a preferred embodiment of the present invention,
3 is a flowchart illustrating a collection procedure of a packet collection unit in a voice-to-text conversion voice recording system according to a preferred embodiment of the present invention, and FIG.
FIG. 4 is a flowchart illustrating a voice segment storage procedure of a speech separator in the speech-to-text conversion transcription recording system according to the preferred embodiment of the present invention.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the drawings, the same reference numerals are used to designate the same or similar components throughout the drawings. In the following description of the present invention, a detailed description of known functions and configurations incorporated herein will be omitted when it may make the subject matter of the present invention rather unclear. In addition, the preferred embodiments of the present invention will be described below, but it is needless to say that the technical idea of the present invention is not limited thereto and can be variously modified by those skilled in the art.

FIG. 1 is a conceptual diagram of a voice-to-text conversion transcription recording system 100 according to a preferred embodiment of the present invention. FIG. 2 is a block diagram of a system for generating a voice content-to-text conversion transcription record 100 according to a preferred embodiment of the present invention. FIG. 3 is a flowchart showing the collection procedure of the packet collecting unit 20 in the voice-to-text conversion transcription recording system 100 according to the preferred embodiment of the present invention, and FIG. Is a flowchart illustrating a voice segment storage procedure of the voice separation unit 30 in the voice-to-text conversion voice recording system 100 according to the preferred embodiment of the present invention.

1, the system for generating a conversation contents voice-to-text conversion transcription according to a preferred embodiment of the present invention includes a server 10, a packet collecting unit 20, a voice separating unit 30, (40), and a text integration unit (50). It is possible to improve the accuracy of voice recognition of conversation contents of a sender / receiver by using Internet telephony (VoIP) and to create a transcription record. It is possible.

Hereinafter, the server 10 will be described in detail.

The server 10 is for a consultant who provides a consultation service using an Internet telephone (VoIP) and a consultant who requests a consultation service. The server 10 includes a DB 11 for storing, managing, and providing call details of the consultant and a consultant .

The DB 11 stores a voice file (MP3, etc.), voice packet, voice data, and the like in which the consultation contents are recorded. A standard language is stored in order to convert voice to text. .

In particular, the administrator can access the server 10 to check words that do not match the standard language stored in the DB 11, modify them into standard words and store them newly, It is possible to automatically set the language to be converted into a standard word, and to add the standard word to the DB 11 in case of a buzzword.

The packet collecting unit 20 is for collecting voice packets and separately collects voice packets of the counselor (agent) and voice packets of the service requester (counselor).

More specifically, in a VoIP call, a voice call is started through a SIP (Session Initiation Protocol) protocol, and actual voice data is transmitted and received through an RTP protocol. The actual voice transmitted and received through the Real-time Transport Protocol (RTP) is digitized by the voice codec at the caller side and divided into payloads of the RTP packet. On the contrary, at the receiver side, In particular, it is possible to recombine the consecutive voice data (complete message) through the value of the Synchronization Source Identifier (SSRC ID) field of the RTP packet header at the receiving end. The combining order of the RTP packets is determined by a sequence number and a time stamp, and the packets of the sender and the receiver are transmitted on the basis of the Synchronization Source Identifier (SSRC ID) value of the RTP packet Store (collect), and store the call start time.

Also, referring to FIG. 2, the order of combining RTP packets is determined by a sequence number and a time stamp. Therefore, voice data (voice packet) of the sender and the receiver (consultant and agent) is separated and stored through the RTP protocol.

Also, referring to FIG. 3, when the VoIP call is started, the RTP packet is collected in the server, the SSRC ID of the packet header is different from the ID of the agent and the consultant, and the RTP packet mirroring (copying) is performed. The mirrored packet separates the RTP packet of the agent and the RTP packet of the consultant based on the SSRC ID, and stores the collected RTP packet in the DB. However, since the mirrored packet of the other agent or the consultant may be introduced at the time of mirroring the RTP packet at the server, If it is not the SSRC ID of the consultant and the agent to be collected, discard the packet.

The voice separation unit 30 stores the voice data separated in units of segments on the basis of the silence period in the voice packets collected by the packet collection unit 20 in the DB 11 of the server 10, A segment time stamp indicating the start point of the data is applied.

In addition, when voice data is extracted from the payload of the RTP packet of the voice separation unit 30, the voice data is divided into segments by separating the segments into segments based on the silence period. The time stamps of the RTP packets A time stamp is generated and applied to each of the segmental audio data.

In addition, when the voice call contents are divided into segments, a time stamp is recorded for each segment based on the start time of the VoIP call and the timestamp information of the RTP (Real-Time Transport Protocol) packet, It is used as a reference value for thermal alignment.

More specifically, referring to FIG. 4, the RTP packets collected in the packet collecting unit are input in sequence number (Sequence Number) order, the voice segment is stored only in the last packet, and the packet excluding the last packet The speech data of a new speech segment unit is generated by separating the speech segments based on the silence period, and time stamps are applied to each of them. If there is no silence interval, it is classified as the last packet and stored as a voice segment.

For example, during a conversation of a separated agent, a call voice of "Your customer's mobile phone charge has been paid (silent section, ie, conversation with the other party) is 000 won" (Copied) into a plurality of RTP packets and stored in the DB. These packets are input into the RTP packet header in the order of sequence numbers (sequence numbers) of the RTP packet headers. Based on the silent interval, ) Are recombined to generate a segment. The segment stores the speech data up to the present when the silent interval (peer talk interval) is found in the packet, and then the packets are generated as a new segment. By repeating the process of rearranging the voice data on a segment basis on the basis of the silence section, the contents of the call can be divided into sentences. In other words, the sentence of the example is separated into two segments separated by the sentence of "Your customer's cell phone charge is not paid" and "Your unpaid amount is 000 won", and the timestamp of the RTP packet header is referred to The segment time stamp is also calculated and stored.

The text conversion unit 40 converts the segmented speech data separated by the speech segmentation unit 30 into text.

The text converting unit 40 may include a noise removing unit 41 and a data filtering unit 42.

More specifically, after the text conversion unit 40 recombines audio data originally transmitted from the stored transmission / reception RTP packets to convert it into text, the audio data MP3 is reproduced to extract a frequency, (41) removes the external noise except for a specific section in which the human voice is irradiated in the frequency domain. Particularly, the recognition and removal of voice and noise sections can be canceled in various ways using a spectrum subtraction method or an LMS algorithm adaptive filter.

Thereafter, the voice data from which the noise has been removed is separated into the voice data of the segment unit, and then the voice data is converted into each text. Here, the text conversion can utilize a technology that is conventionally used, such as Google's speech recognition engine or a speech recognition device of a mobile device.

In particular, the voice packets of the collected songs and receivers are divided and stored in units of segments on the basis of a silent section, so that the text conversion section 40 separates voice packets based on a silent section, The recognition accuracy can be improved.

Thereafter, the accuracy of the text conversion is further improved by using the data filtering unit 42 of the text conversion unit 40.

More specifically, if the converted text including the recognition error word, buzzword, and dialect does not match the standard word and sentence interval of the server 10 DB 11, it is checked and stored in the DB 11 do.

In the case of a standard word stored in the DB 11 or a text that does not coincide with the sentence, the administrator accesses the server 10 and registers it in a new standard word or standard sentence, Converted into newly registered words and sentences, and filtered, thereby improving the accuracy of recognition sentences.

For example, assuming that the text converted by the text conversion unit 40 is 'paper money', 'reserved', and 'received day', it is converted into text by segmenting on a segment basis based on a silent section. If the server 10 determines that the 'reserved' among the standard words stored in the server 10 DB 11 is that there is no standard word or sentence matching the 'match', ' The server 10 checks whether the user has accessed the server 10, or if the administrator accesses the server 10, the user can register the new information with a standard word or sentence, Is automatically converted into a modified or newly registered word or sentence.

The text integrating unit 50 integrates the converted text in the text converting unit 40 and stores it in the DB 11. [

In addition, the text integrating unit 50 arranges the sentences of the consultant and the consultant who are speech-recognized on a segment-by-segment basis and converted into text into a time series based on the segment time stamp to generate a transcript.

In more detail, the text integrating unit 50 can generate a transcript created in time sequence by arranging transmission / reception sentences converted in units of segments on a time-series basis based on a segment time stamp, It is easier to search for words in the future when checking the transcripts, and it is faster to store them in a text file than to store the voice files, so that the speed is faster and a lot of data is stored in the DB 11 Of course.

It will be apparent to those skilled in the art that various modifications, substitutions and substitutions are possible, without departing from the scope and spirit of the invention as disclosed in the accompanying claims. will be. Therefore, the embodiments disclosed in the present invention and the accompanying drawings are intended to illustrate and not to limit the technical spirit of the present invention, and the scope of the technical idea of the present invention is not limited by these embodiments and the accompanying drawings . The scope of protection of the present invention should be construed according to the following claims, and all technical ideas within the scope of equivalents should be construed as falling within the scope of the present invention.

10-server
11 - DB
20 - packet collecting unit
30 -
40 - Text conversion section
41 - Noise canceling
42 - Data filtering unit
50 - text integration unit
100 - Voice-to-speech transcription recording system

Claims (3)

A server for a counselor who provides counseling service using VoIP and a counselor who requests counseling service and includes a DB for storing, managing, and providing call history of the counselor and the counselor;
A packet collecting part for collecting voice packets, the packet collecting part collecting voice packets of a consultant separately from voice packets of a service requester;
A voice separator for storing voice data obtained by dividing voice packets collected by the packet collecting unit into segments;
A text conversion unit for converting the speech data of the segment unit separated by the speech separation unit into text; And
A text integration unit for integrating the converted text in the text conversion unit and storing the integrated text in a DB
Lt; / RTI >
The speech separator
When the voice packets collected by the packet collecting unit are divided into segments, the packets collected in the packet collecting unit are input in order of the order of the numbers, and only the last packet is stored as voice segments. Of the voice packets excluding the last packet, And stores them as voice segments. If there is no silence interval, they are classified as last packets and stored as voice segments
/ RTI >
The speech separator
Segment timestamps are applied to the segmented data, which shows the viewpoint of the data.
/ RTI >
The text conversion unit
A noise removing unit for removing noise excluding a specific section in which a human voice is extracted by extracting a frequency by reproducing voice data,
A data filtering unit for connecting to the server and registering or modifying a new word or a sentence when the converted text does not match the standard word or sentence stored in the DB,
Further comprising: a voice-to-text conversion transcription recording system. delete delete

Images