KR100216536B1 - Automatic public discontent processing method - Google Patents

Abstract

The present invention relates to an interactive automatic complaints processing method in which a civil servant automatically issues a civil affairs document through a public telephone network, as well as guide services related to various civil affairs procedures.

Interactive automatic complaint processing method according to the present invention for achieving the above object is a first step of the PC manager and the signal processing manager to initialize the entire system to start the civil service, and the signal processing manager is ringing A second step of guiding the complainant and greetings to the complainant, and the signal processing manager inducing a complaint content and an outgoing phone number of the complainant, processing the DTMF signal, and recognizing a word of the complainant's voice; and And a fifth step of guiding and returning the receipt number and greeting to the complainant after the word recognition of the complainant voice.

Therefore, by establishing a 24-hour administrative guide and reception system for both the day and the night where the employees of the government office are absent, not only can the unnecessary waiting time of civil complaints be eliminated, but also the efficiency of work performance by automating the issuing business can be improved. It can also be used as a countermeasure to improve the existing postal complaint system.

Description

Interactive automated complaint handling method

1 is a block diagram showing a hardware specification of the present invention.

2 is a flowchart showing a processing procedure for the interactive automatic complaint processing method according to the present invention.

The present invention relates to an interactive automatic complaint processing method in which a civil servant automatically issues a civil affairs document of a government office through a public telephone network as well as guide services related to various civil affairs.

In general, there are three methods of handling complaints using a telephone: a method in which a person makes a phone call and communicates directly with a person handling a complaint in a government office, a method using an audio response system (ARS), a personal computer ( PC) and the like.

First, the first method is for a complaint person to directly call the relevant complaint handling person in the public office, receive the relevant documents through communication with the person in charge, and the complaint person visits the public office and issues the desired documents. However, there is a disadvantage that the civil service is not available when the person in charge of the civil service is absent or when an emergency arises, and the service is not immediately and flexible due to the limited communication line.

Secondly, the ARS method, in which a complainant receives related documents according to a certain induction schedule presented by a computer through a telephone line, has a certain order according to a pre-recorded mechanical broadcast procedure, which causes a sense of rejection. In the case of long or mistyped inputs, it takes a lot of time, such as having to repeat the whole process from the beginning, and it is easy to feel rejected because it lacks the naturalness and flexibility of communication.

Finally, through the modem connected between personal computers (PCs), you can access the service computer system of the public telephone network company, view the contents of the complaints presented on the computer screen, and select the corresponding complaints with the keyboard. PCs are expensive, and they must run software systems that require specific expertise, which is still an obstacle for people who are not familiar with PCs.

The present invention has been made in order to solve the above problems, an interactive automatic civil complaint processing to analyze the requirements of civil complaints coming through the public telephone network to automate document reception and issuance for administrative guidance and issuance of civil documents The purpose is to provide a method.

Interactive automatic complaint processing method according to the present invention for achieving the above object is a first step of the PC manager and the signal processing manager to initialize the entire system to start the civil service, and the signal processing manager is ringing Recognizing the second step of guiding the greetings and how to use the complaints, the signal processing manager derives the complaints and the outgoing phone number of the complaints DTMF signal processing and the third step of extracting the features of the complaints, the PC administrator And a fifth step of recognizing the words of the complaint voice and a fifth step of guiding and returning the reception number and the greeting to the complaint after the word recognition of the complaint voice.

Hereinafter, an interactive automatic complaint processing method according to the present invention will be described in detail with reference to the accompanying drawings.

First, as shown in FIG. 1, according to the hardware specification of the present invention, it can be implemented on a personal computer (PC) compatible with the most widely used IBM-PC and MS-DOS. It can be realized in all models equipped with central processing unit and main memory device of 8MByte or more.

In addition, up to four boards can be controlled on the standard communication bus of a PC by the Digital Signal Processing (DSP) board, which is responsible for executing voice signal processing, dialogue control and subroutines for voice feature extraction. The communication between each signal processing manager and the PC system uses the standard address method of the RS-232 card, which is the standard communication interface of the PC, so that the peripheral device standard interface is applied. Allocates communication ports of COM2, COM3, and COM4.

Each signal processing manager generates an interrupt to the PC's central processing unit to transmit and receive data and control signals or codes with the PC. For this purpose, a standard interrupt request line (IRQ) set in the PC is set. IRQ10, IRQ11, IRQ12, and IRQ15 are assigned to the respective signal processing managers.

Human voice is an analog type of material that is fundamentally continuously changing and must be converted into a digital signal in order to be processed in a computer.

In addition, in order to reproduce the audio data processed digitally, i.e., stored in the memory, it is necessary to convert back to an analog signal.

In addition, the hard disk built into the PC has a vector quantization codebook which is compared with a feature vector calculated during speech recognition, and a reference pattern for each word that is a recognition object as a binary pattern. Broadcasting voice data, such as greetings, are digitally coded and binarized at a resolution of 8KHZ (Sampling Rate), and the user's voice is also encoded and stored at the same resolution when talking with the complainant.

According to the software specification of the present invention, it can be divided into two parts according to the place to be performed, which is a program executed in the central processing unit on the PC side and the dialogue control and voice feature extraction processed by the signal processing manager.

On the PC side and DSP side, there is one manager program for each system management.

That is, the PC manager program controls the whole process of performing feature extraction and recognition of voice by exchanging messages and control codes through communication with each of the four signal processing managers.

In addition, the PC manager program supports virtual real-time multi-processing to serve four complaints simultaneously, and operates virtual memory to overcome the operating limits of MS-DOS main memory. have. This virtual memory device is used to store the reference pattern of the hidden Markov model used for speech recognition. The virtual memory device is divided into a predetermined number, that is, the number of words to be recognized, and then each reference pattern is stored. have. In recognition, the hidden Markov model frequently accesses the virtual memory device, reads the data, and performs the recognition process by comparing with the input pattern.

The present invention is described in more detail as follows.

As shown in the flow chart showing the processing procedure for the interactive automatic complaint processing method of FIG. 2, first, the PC administrator simply initializes the graphic user interface which outputs only the current operation state of the system to the screen (S10). HMM model parameter for recognizing the HMM (Hidden Markov Model) stored in the form of a file from the hard disk after performing the proper partitioning of the virtual memory device as described above. Initializing the virtual memory device storing the parameters in the virtual memory device (S20).

Here, the HMM is a statistical pattern recognizer, also called a hidden Markov model, which explicitly models the statistical characteristics of the speech signal by means of the Markov processor and its stochastic states. That is, the speech recognition using the HMM in the present invention uses the index of the codebook extracted through three multiple codebooks used as the reference pattern of the recognizer for each word using the Baum-Welch method. It is designed to extract the probability statistical parameters, and it is composed of learning to determine the recognition words by maximizing the statistical likelihood through the Viterbi search based on the extracted parameters.

In general, this statistical process is implemented by a finite state automata.

Subsequently, the signal processing manager is initially initialized by the PC manager. First, the PC manager creates an environment required for communication between the PC and the signal processing manager by initializing the DSP driving device (S30).

At this time, a system definition and an environment regarding interrupt signals, channels, data buffers, and the like with the DSP manager are set (S40, S50).

The interrupt signal sets IRQ10, IRQ11, IRQ12, and IRQ15 in the order of COM1, COM2, COM3, and COM4, and the channel numbers are similarly defined from 0 to 3 as the serial numbers.

In addition, the buffer to be used in the channel is set to be used independently.

This completes the system definition and configuration.

It then communicates with the signal processing board via the bus using an interrupt scheme that defines a signal processing manager program.

The signal processing board then receives the manager program and plays the role of manager.

At this time, the signal processing manager and the PC establish a system definition and environment such as establishing a channel by exchanging a communication-only message that communicates with each other (S40 ', S50').

As such, when channel establishment, system definition, and environment establishment are completed, PC and DSP managers set the environment necessary for program execution such as defining and allocating temporary or static memory required by both sides (S60, S60 ').

Subsequently, the PC manager enters a wait state (interrupt wait state) waiting for a command of the signal processing manager (S70). In this case, the signal processing managers are always required to send a command and receive a result using an interrupt scheme if necessary.

At the same time, the signal processing manager enters the standby state to receive a call from the civil affairs after the environment setting is completed, similarly to the interrupt standby state (S80).

In this case, the standby state means detection of a call ringing from the complaint, and upon being detected, the standby state ends and the service person enters the service phase.

When DTMF (Dual Tone Multi Frequency), that is, a telephone ringing, the signal processing manager detects this and enters a conversation service state for message transmission and reception between users (S90 and S100).

In other words, the user broadcasts a greeting and a brief announcement to the user, and transmits to the PC administrator the current state, for example, a call from a civilian to start the service as it is through the channel operation block (S110).

At this time, the signal processing manager receives two types of signals from the complainant, one of which is the voice of the complainant and the other of which is the number sound generated by the complainant's phone.

This means that the service to the complainant is performed in two ways: voice recognition and DTMF signal processing.

Once in a conversation with the complainant, the signal processing manager requests the complainant to speak the necessary words or press the number plate of the phone. The complainant then speaks the word or press the number plate of the phone at that time (S160).

Then, the signal processing manager records this (S170) and then performs DTMF signal processing and extraction of voice characteristics (S180 and S190).

Then, after extracting the voice characteristics and matching the codebook, the signal processing manager transmits the interrupt list to the PC manager along with the interrupt code to the PC manager.

At this time, the PC manager who receives the interrupt interprets the interrupt code (S130), receives the voice characteristic, calls the hidden Markov performing module, performs the recognition process using the same (S140), and then presents the recognition result to the PC manager. The PC manager sends the signal back to the signal processing manager (S150) to inform the complainant of the recognition result and goes through the procedure of re-evaluating whether the recognition is accurate.

In addition, the PC manager notifies that the recognition has been completed, sends a return value indicating the execution result (S220), and returns to the interrupt waiting state again.

The complainant only needs to answer yes or no.

All these processes are sent to the PC manager every time to output the progress in the form of a text on the screen, and when the conversation with the complaint ends and broadcasts the termination message and greeting and reception number and hangs up the telephone line (S200), The phone enters the ring detection waiting state again (S210).

As described above, the interactive automatic complaint processing method according to the present invention establishes a 24-hour public administration and reception system for both the day and the night where the employees of the public office are absent, eliminating unnecessary waiting time of civil complaints and automating the issuing work. Not only can it be efficient in its performance, but it can also be used as a countermeasure to improve the existing postal complaint system provided in isolated areas.

In addition, the complainant can have the same expected effect as having the terminal by talking to the computer through voice.

In addition, the current system is likely to contribute to the improvement of public welfare by enabling one-time visit service and automatic civil service service when it is linked with facsimile already distributed to each institution, company and home.

Claims (1)

In the first step, a PC manager and a signal processing manager initialize an interface, a virtual memory device, a DSP drive device, and a storage location to define a system and set up an environment to initialize the entire system in order to start a complaint service. Recognizing the caller's phone ring, the second step of guiding the greeting and how to use the complaint, the signal processing manager derives the complaint content and the outgoing phone number of the complaint, DTMF signal processing, and the third step of extracting the features of the complaint voice The PC manager performs a fourth step of recognizing the word of the complaint voice and a fifth step of transmitting a receipt number and a greeting to the complaint person after recognizing the word of the complaint voice. Interactive automatic complaints processing method characterized in that the processing of the complaints automatically while talking.