Development and Application of Automatic Speech Recognition Technology (ASR) in Computer Aided Teaching in Deaf Children
Rdte of Cai for the Deaf & Dumb Children Based on ASR
I. The purpose and significance of the source and research of the topic;
According to relevant institutional sampling surveys, there are about 60 million persons with disabilities, including about 1 million deaf people, and the deaf people under 18 years old should be more than 1 million, which is a huge vulnerable group. They have a lot of hardships and obstacles that are difficult to imagine in learning, work and life, and have a multiplier effort in the path of growth and development. The education of deaf childhood is a glorious business that is full of love, full of socialist humanity. It is also our entire social righteous responsibility.
Some economic developed countries and regions such as Europe, United States, and Hong Kong were very early, and the level was high. In traditional teaching methods, teaching methods, instruments and other aspects, my country is larger. At this stage, domestic and exemption education, rehabilitation agencies are paying attention to the research and equipment of special equipment, and in the application of modern teaching technologies, especially in the application of the latest scientific and technological achievements such as network, computer, automatic voice recognition technology, my country has existed A certain gap is basically in the infancy. In fact, modern information and new theories, new technologies, etc. in the fields of other science and technology, and new equipment will be significantly changed by the concept of deaf rehabilitation, education, etc., which will will change the concept of the world. The education reform brings a huge impact, and my country's deaf education must seize opportunities, in-depth reform and innovation, and strive to improve the quality of education in schools and keep up with the pace of the times.
In the educational field of deaf children around the world, from the application of modern educational technology, my country and foreign gaps are not large, this is a rare opportunity to catch a world level of the world. This topic is the first to combine modern information technology, modern educational technology and special demand for education. By applying information technology and software development achievements in deaf education, and summarizes the implementation of objects, implementation methods and effects, compliance The Implementation Experience of Modern Education Technology of China 's National Affairs in Deaf Children's Education.
This topic aims to use modern educational technology and means based on automatic speech recognition technology (ASR), in the technical level of deaf people, research, reform, practice, and pay attention to cultivating deaf children's learning, so that they collect, Analysis of the ability to analyze and deal with information, master the modern means of self-study and lifelong learning, improve their own quality and adapt to social life, so that it can participate in social life with equal position and equal opportunity, share social material cultural achievements.
Second, the research status and analysis of domestic and foreign in this direction;
The study of deaf-mute voice training foreign countries began in the middle of the 1960s, depending on the feedback pathway, can be roughly divided into two categories: using the deaf people's residual hearing with the hearing aid to listen to the audio and correct pronunciation of the pronunciation, the auditory feedback training system and pass The visual feedback training system is correctly corrected based on the characteristic parameters obtained after the processed by its own pronunciation. The former is low, but the effect is poor. Listening, severe deafness, and the patient's patient effect is poor or completely invalid. The latter is almost applicable to all deaf people because of visual feedback, and the training effect is better. At the early stage, the visual feedback system is high, with the development of computer and large-scale integrated circuit technology, especially the emergence of voice-specific chips and single chip, the cost has been greatly reduced.
Currently, there have been a variety of visual auxiliary voice training systems. These systems are basically processed by the trainee, and extract the characteristics of speech (eg, strength, duration, spectrum, base frequency, resonance peak, etc.), and the characteristics of the standard pronunciation are displayed on the CRT, let The trainee compared to its own pronunciation with standard sounds, step by step to correct your pronunciation. Unfortunately, the information displayed by this system is too transformed into a general trainee. It is not easy for them, especially deaf children, so affecting the training effect. This is the biggest shortcoming of such systems. As for the focus of this topic, automatic speech recognition research begins in the early 1950s, the electronic signal spectrum analysis instrument began to be used to identify simple, small amounts of syllables and phonemes from speech signals. With the rapid development of computer technology, after entering the 1990s, the study of speech recognition further warmed, in addition to continuous voice listening, there have been many practical research directions. The ViaVoice, which is the first to launch the first ViaVoice, non-specific people, continuous speech recognition technology is tending to mature. At present, there are still many mature speech ASR products on the market, and most of them support secondary development, such as Microsoft's Speech Application SDK (SASDK), SUN advocated Javaspeechapi, IBM DUTTY , etc.. Most of them can identify English, Japanese and Chinese languages, Dutty , can even identify dialects in certain areas, such as Guangdong's dialects - Cantonese. From the prospect of the development of the entire speech recognition research, robust (Robust) of the speech recognition system will be one of the focus of the next few years. Because this is the most urgent key issue that the speech recognition system is converted from the experiment to the utility process. The automatic speech recognition system for the deaf boy will be a key to the research that cannot be ignored.
my country's speech recognition research is late, but due to the increasing importance of Chinese speech recognition, the development of the last decade has been very rapid. Therefore, the automatic speech recognition of the beginning of the 1990s can basically be synchronized with abroad. At present, the research institutions in this area mainly include: Tsinghua University, Academy of Academy of Chinese Academy of Sciences, Institute of Automation, Chinese Academy of Sciences, University of Science and Technology, China University of Science and Technology, University of Defense Science and Technology, Beijing University of Posts and Telecommunications and so on.
Third, the main research content;
ASR, English full name is Automated Speech Recognition, an automatic speech recognition technology, which is a technology that converts people's voice into text. Voice recognition is a multi-discipline cross-connected area, which is closely linked to acoustic, phonology, linguistics, digital signal processing theory, informationism, computer science. ASR against the education of the child will also be used to include many disciplines including education and education technology.
This topic is based on speech recognition technology, which is to process the trainee, calculate the approximate rate of standard speech and adjacent speech, and display its result on the CRT to guide the trainee's pronunciation to the standard voice. The system displays an approximation rate with standard speech and other similar sounds, and complements the pronunciation of the mouthpiece, for the trainee, the display information is intuitive, and the training effect is good.
Specifically, this topic focuses on the development of automatic speech recognition system applicable in computer-aided teaching of deaf-mute children, namely small (specific) vocabulary, non-specific deaf boy, continuous adaptive speech recognition system, and provides convenient front desk (computer Auxiliary teaching platform) calling interface.
Fourth, research programs and progress arrangements, the goal is expected;
Research methods and technology routes intended to be taken (including overall arrangements, steps and progress, etc.):
This topic is based on Java language, combined with the algorithm principle of automatic voice recognition technology. Take the following way:
Stage 1: Analysis stage, study many academic achievements, identify specific algorithms
January 2005 to February 2005
Stage 2: Design phase, design flow chart.
February 2005 to March 2005
Stage 3: Development procedures, according to the design phase process, encoding.
March 2005 to April 2005
Stage 4: Test Maintenance Procedure, perform practical tests according to the actual situation.
April 2005 to May 2005
Stage 5: Realize product input, complete graduation design tasks.
The 1st to 3 stages are not clearly defined, screw development
The results of the study and final results:
Staged results:
By January 2005, submit a feasibility investigation report;
By March 2005, the design of the flowchart is completed;
By April 2005, the coding is completed;
By May 2005, the software test was completed, and the final scientific research results were submitted.
5. Conditions and funds required for completion of the topic;
1. Already have the conditions: ASR's theoretical research results, related JSAPI documents, etc.
2. Required conditions: Developing a PC Computer Aided Teaching Platform
3. Funding issues: In actual development, considering copyright issues, you may need to purchase some related software, or consulting fees.
Sixth, it is expected that difficulties and problems that may encounter during research and solutions;
1. Theory
Question: Since the theory of speech recognition is quite mature, the discipline involved is extensive, so I cannot solve some ways.
Solution: Consult the relevant expert or consult the relevant information
2. Technique
Question: Now comparative development language is C / C , and this topic is Java, there may be few reference materials, difficult development
Solution: A series of Javaspeechapi that has been released and perfect
Seven, main references.
[1] Chen Yizhen et al. Demoniology, deaf-minded visual counseling speech training system, China, China, China Biomedical Engineering, China
[2] Li Jianmin et al. Journal of Practical Identification System, Based on Chinese Voice Characteristics 1992.5
[3] Hao Jie Based on Classic Hidden Markov Model of Chinese Continuous Voice Recognition System Electronic and Information 2002.7
[4] Wang Yue Speech Identification Adaptive Technology Research and Realization Master's thesis 2000.5
[5] Research and Basic Realization Master of Lei Jing Speech Recognition Technology 2002.3.1
[6] Master's degree in Research and Application of Chen Liolong Continuous Voice Recognition 2002.3.1
[7] Wang Zhiqiang Based on GMM's Sound Signal Classifier Research Master's Paper 2003.6.30
[8] Master's degree in Design and Implementation of Automatic Voice Response Systems in Cao Zhi Wall 2004.2.1
[9] Wang Yue Based on HMM Model Master of Embedded Speech Recognition Software 2003.3.1
[10] Zhang Jun's research on anti-noise sound recognition technology for research on doctorate 2003.5.1
[11] Wang Ning based on phoneme-specific people's big vocabulary Chinese speech recognition algorithm research Master's thesis
[12] Shengqing Voice Automatic Identification Technology (ASR) and Master of Real Time Real Time Software 2001.3.1
[13] Sun Java Speech API PROGRAMER's Guide Development Guide 1998.10.26
