Research on Voiceprint recognition method of buried drainage pipe based on MFCC and GMM-HMM

An voiceprint recognition model based on short-term energy and Gaussian Mixed Hidden Markov models was proposed in this paper aiming to solve the problem that the number and position of the of the blockages, pipe components and pipe end in the buried drainage pipeline are uncertain and difficult to identify. This method first performs wavelet threshold de-noising method on the sound pressure signal collected in the operating pipeline, and then uses the short-term energy as the discriminant parameter for endpoint detection, then divide into segmentations obtain the sound pressure signal corresponding to different types of individuals in the pipeline, and then extracts the Mel-frequency Cepstrum coefficient to constructs feature vector sets that can characterize the blockage, lateral connection and the pipe end, and finally the feature vector sets are used to train the Gaussian Mixed hidden Markov model to obtain the classification results. The experimental results have shown that the proposed method can effectively identify blockages, pipe fittings such as lateral connection and pipe end, furthermore, it avoids the problems of low recognition rate of new samples caused by insufficient known conditions included in the training samples.