Study On Speech Signal Processing Base On Speech Rehabilitation Training System For Autism

People have paid more and more attention to rehabilitation treatment of autism, andspeech training is the essential link in rehabilitation training of autism. The traditionalrehabilitation training need medical personnel or relatives involved, especially need toconduct oral repetition training, so the great mental pressure and economic burden onfamily to involve rehabilitation training. With the development of computer technology,especially the speech processing technology matures, in the field of medicine, thetechnology which aided by computer for various rehabilitation therapies has become animportant research direction. A new method for speech rehabilitation training for childrenwith autism is proposed in this thesis. And the method is made by the computer aided inspeech rehabilitation training. This thesis focuses on developing an intelligential system ofspeech rehabilitation training for children with autism, which is flexible and extensible.The principle and method of system is discussed in detail. And the technology anddesigning of system software is introduced in this thesis.The main work of the thesis includes: designed the overall program of speechrehabilitation training system by drawing from the basic method. And the way of signalacquisition, data preprocessing, noise removal of data, which is based on the technology ofvoice processing and analysis. Then it is designed a method of endpoint detection andsyllable segmentation, which is based on the technology of short-time autocorrelationenergy and short-time zero spectral product. Moreover, it is detected the characteristicvalue of voice, e.g. pitch period, formant, LPCC, MFCC. Then the method of Hopfiledneural network combination with BP artificial neural network was proposed to used forspeech recognition. The Hopfiled neural network was used for associative memory, so asto make for the change trend both voice characteristic values of recognition andcharacteristic values of the training sample to be consistent, then using the BP network torecognize.In this procedure, the combination artificial neural network was used for apreliminary study on the speech recognition of children with autism.On the basis of these studies, it is designed software of rehabilitation training system.And it is realized the biological feedback training by giving people with autism to providevoice, text, images and other forms of sound and image signals. The Software of thesystem makes training course easy to expand, because it combined technology of text-to-speech (TTS) and used computer synthesized speech as a voice training course.