An Improved Speech Segmentation and Clustering Algorithm Based on SOM and K-Means

This paper studies the segmentation and clustering of speaker speech. In order to improve the accuracy of speech endpoint detection, the traditional double-threshold short-time average zero-crossing rate is replaced by a better spectrum centroid feature, and the local maxima of the statistical feature sequence histogram are used to select the threshold, and a new speech endpoint detection algorithm is proposed. Compared with the traditional double-threshold algorithm, it effectively improves the detection accuracy and antinoise in low SNR. The k-means algorithm of conventional clustering needs to give the number of clusters in advance and is greatly affected by the choice of initial cluster centers. At the same time, the self-organizing neural network algorithm converges slowly and cannot provide accurate clustering information. An improved k-means speaker clustering algorithm based on self-organizing neural network is proposed. The number of clusters is predicted by the winning situation of the competitive neurons in the trained network, and the weights of the neurons are used as the initial cluster centers of the k-means algorithm. The experimental results of multiperson mixed speech segmentation show that the proposed algorithm can effectively improve the accuracy of speech clustering and make up for the shortcomings of the k-means algorithm and self-organizing neural network algorithm.

Research on Snoring Recognition Algorithms

A snoring recognition algorithm based on machine learning is proposed to effectively and precisely recognize snoring. To obtain a dataset, the speech endpoint detection algorithm and Mel frequency cepstrum coefficient feature extraction algorithm are applied to process speech signal samples. The dataset is classified into snoring and nonsnoring data (other speech signals) using support vector machines. Experimental results show that the algorithm recognizes snoring signals with a high accuracy rate of 97% and positively impacts subsequent research and related engineering applications.

LTSD and GDMD features for Telephone Speech Endpoint Detection

This paper proposes a new contour-based speech endpoint detector which combines the log-Group Delay Mean-Delta (log-GDMD) feature, an adaptive twothreshold scheme and an eight-state automaton. The adaptive thresholds scheme uses two pairs of thresholds - for the starting and for the ending points, respectively. Each pair of thresholds is calculated by using the contour characteristics in the corresponded region of the utterance. The experimental results have shown that the proposed detector demonstrates better performance compared to the Long-Term Spectral Divergence (LTSD) one in terms of endpoint accuracy. Additional fixed-text speaker verification tests with short phrases of telephone speech based on the Dynamic Time Warping (DTW) and left-to-right Hidden Markov Model (HMM) frameworks confirm the improvements of the verification rate due to the better endpoint accuracy.