The Characteristic Parameters Of Pattern Matching In Voice Score

In the language learing, the inaccurate pronunciation becomes a inartificial barriers to learn non-native tongue.In the paper, the speech evaluation is based on the standard speech. It is a method that evaluates the similarity between a test speech and the standard speech. When the test speech resembles the standard more, the socre will be higher. We use various approaches for speech feature extraction, pattern matching, and similarity computation. In particular, we use magnitude contour, pitch contour, and mel-frequency cepstral coefficients as the features to generate a similarity score.The main achievements are listed as follows:1. Endpoint detection algorithm, aiming at the shortness of the traditional double-threshold endpoint detection algorithm in noisy environment, some improvements is given. Especially in dealing with the actual speech, through many trials and statistics calculate the threshold of the average short-term energy and zero-crossing rate from the noise speech.2. Feature extraction algorithm, several common feature paramenters for speech recognition is systematically summarized. The principle, implementation details, advantages and disadvantages of feature based on Linear Prediction Coding Coefficient (LPCC) and Mel Frequency Cepstral Coefficient (MFCC) are analyzed in detail, aiming at the problems of the MFCC feature, the wavelet transform algorithm have been given. In order to improve the boustness of the systems, noisy speech is decomposed into various frequency bands and then de-nosing is processed by TEO in every frequency band, then the output of various frequency bands through wavelet reconstruction to restore the signal. Finalyy, wavelet coefficient is transformed into MFCC.3. Speech evaluation mechanisms, through the processing of a complete English sentence can gain the mechanisms of the speech evaluation. The speech evaluation system is based on the speech processing, respectively, take pre-processing to the standard and the test speech, and the feature extraction, feature warping paramenters and pattern matching etc. to arrive the three arameters of similarity and parameters by studying the score in the speech mechanisms in the proportion of the establishment of a suitable rating.