The Reserch On Recognition Of Aphasia Speech Signals

In recent years, with the rapid development of computers and speech recognition technology, the rehabilitation of aphasic patients has made significant progress. The essence of adjuvant therapy on patients is to analysis and process voice signals and finally achieves its purpose of identification. The key technology is the identification of patients with aphasia speech signal. The speech recognition is the current hot of aphasia patients in rehabilitation therapy system of patients with aphasia today. The main topics of this paper are as follow:First, the background of the topics, significance and main contents have been described, with emphasis on current aphasia, speech recognition technology research.Secondly, the speech recognition technology knowledge has been studied. According to the characteristics of speech signal in patients with aphasia, speech signal recognition principle was clarified and made the speech recognition program in patients with aphasia.Third, built a test platform for speech acquisition, completed its acquisition of speech signals in patients with aphasia as the voice behind the voice signal processing and identification of the original data.Fourth, a new wavelet decomposition structure - detailed high frequency components, combined with soft thresholding algorithm of speech signal denoising was achieved for patients with noise polluted voice signal. Experiments show that this method achieved good denoising effect to meet the requirements of keeping more real signals.Fifth, the voice signal of aphasia patients has intermittent characteristics, thus, conventional endpoint detection and feature extraction methods have been explained. With theoretical analysis, a dual threshold endpoint detection method and MFCC feature extraction for follow-up was used to match the data provided.Finally, with the comparison of multiple speech recognition algorithms, studied the speech recognition algorithm according to papers dealing with the specific requirements of voice signals by analyzing and selecting the DTW algorithm, and completed the identification of patients with aphasia speech signal. Theoretical analysis and experimental simulation results show the effectiveness of the proposed method.