A parameter estimation algorithm of the continuous hidden Markov model isintroduced and the rigorous proof of its convergence is also included. The algorithm uses theViterbi algorithm instead of K-means clustering use...A parameter estimation algorithm of the continuous hidden Markov model isintroduced and the rigorous proof of its convergence is also included. The algorithm uses theViterbi algorithm instead of K-means clustering used in the segmental K-means algorithm to determineoptimal state and branch sequences. Based on the optimal sequence, parameters are estimated withmaximum-likelihood as objective functions. Comparisons with the traditional Baum-Welch and segmentalK-means algorithms on various aspects, such as optimal objectives and fundamentals, are made. Allthree algorithms are applied to face recognition. Results indicate that the proposed algorithm canreduce training time with comparable recognition rate and it is least sensitive to the training set.So its average performance exceeds the other two.展开更多
As a kind of statistical method, the technique of Hidden Markov Model （HMM） is widely used for speech recognition. In order to train the HMM to be more effective with much less amount of data, the Subspace Distribut...As a kind of statistical method, the technique of Hidden Markov Model （HMM） is widely used for speech recognition. In order to train the HMM to be more effective with much less amount of data, the Subspace Distribution Clustering Hidden Markov Model （SDCHMM）, derived from the Continuous Density Hidden Markov Model （CDHMM）, is introduced. With parameter tying, a new method to train SDCHMMs is described. Compared with the conventional training method, an SDCHMM recognizer trained by means of the new method achieves higher accuracy and speed. Experiment results show that the SDCHMM recognizer outperforms the CDHMM recognizer on speech recognition of Chinese digits.展开更多
文摘A parameter estimation algorithm of the continuous hidden Markov model isintroduced and the rigorous proof of its convergence is also included. The algorithm uses theViterbi algorithm instead of K-means clustering used in the segmental K-means algorithm to determineoptimal state and branch sequences. Based on the optimal sequence, parameters are estimated withmaximum-likelihood as objective functions. Comparisons with the traditional Baum-Welch and segmentalK-means algorithms on various aspects, such as optimal objectives and fundamentals, are made. Allthree algorithms are applied to face recognition. Results indicate that the proposed algorithm canreduce training time with comparable recognition rate and it is least sensitive to the training set.So its average performance exceeds the other two.
基金Supported by the National Natural Science Foundation of China (No.60172048)
文摘As a kind of statistical method, the technique of Hidden Markov Model （HMM） is widely used for speech recognition. In order to train the HMM to be more effective with much less amount of data, the Subspace Distribution Clustering Hidden Markov Model （SDCHMM）, derived from the Continuous Density Hidden Markov Model （CDHMM）, is introduced. With parameter tying, a new method to train SDCHMMs is described. Compared with the conventional training method, an SDCHMM recognizer trained by means of the new method achieves higher accuracy and speed. Experiment results show that the SDCHMM recognizer outperforms the CDHMM recognizer on speech recognition of Chinese digits.
