Based on confusions between hidden Markov model (HMM) states, a state-restructuring method was proposed. In the method, HMM states were restructured by sharing Gaussian components with their related states, and the re...Based on confusions between hidden Markov model (HMM) states, a state-restructuring method was proposed. In the method, HMM states were restructured by sharing Gaussian components with their related states, and the re-estimation to the increased-parameters, i.e., the inter-state weights, was derived under the expectation maximization (EM) framework. Experiments were performed on speaker-independent, large vocabulary, continuous Mandarin speech recognition. Experimental results showed that the state-restructured systems outperformed the baseline, and achieve significant improvement on recognition accuracy compared with the conventional parameter-increasing method. Such comparative results confirmed that the state-restructuring method was efficient.展开更多
This paper focuses on a state sharing method for an artificial neural network (ANN) and hidden Markov model (HMM) hybrid on line handwriting recognition system. A modeling precision based distance measure is proposed ...This paper focuses on a state sharing method for an artificial neural network (ANN) and hidden Markov model (HMM) hybrid on line handwriting recognition system. A modeling precision based distance measure is proposed to describe similarity between two ANNs, which are used as HMM state models. Limiting maximum system performance loss, a minimum quantification error aimed hierarchical clustering algorithm is designed to choose the most representative models. The system performance is improved by about 1.5% while saving 40% of the system expense. About 92% of the performance may also be maintained while reducing 70% of system parameters. The suggested method is quite useful for designing pen based interface for various handheld devices.展开更多
Abstract： A hierarchical method for scene analysis in audio sensor networks is proposed. This meth-od consists of two stages： element detection stage and audio scene analysis stage. In the former stage, the basic au...Abstract： A hierarchical method for scene analysis in audio sensor networks is proposed. This meth-od consists of two stages： element detection stage and audio scene analysis stage. In the former stage, the basic audio elements are modeled by the HMM models and trained by enough samples off-line, and we adaptively add or remove basic ele- ment from the targeted element pool according to the time, place and other environment parameters. In the latter stage, a data fusion algorithm is used to combine the sensory information of the same ar-ea, and then, a role-based method is employed to analyze the audio scene based on the fused data. We conduct some experiments to evaluate the per-formance of the proposed method that about 70% audio scenes can be detected correctly by this method. The experiment evaluations demonstrate that our method can achieve satisfactory results.展开更多
文摘Based on confusions between hidden Markov model (HMM) states, a state-restructuring method was proposed. In the method, HMM states were restructured by sharing Gaussian components with their related states, and the re-estimation to the increased-parameters, i.e., the inter-state weights, was derived under the expectation maximization (EM) framework. Experiments were performed on speaker-independent, large vocabulary, continuous Mandarin speech recognition. Experimental results showed that the state-restructured systems outperformed the baseline, and achieve significant improvement on recognition accuracy compared with the conventional parameter-increasing method. Such comparative results confirmed that the state-restructuring method was efficient.
文摘This paper focuses on a state sharing method for an artificial neural network (ANN) and hidden Markov model (HMM) hybrid on line handwriting recognition system. A modeling precision based distance measure is proposed to describe similarity between two ANNs, which are used as HMM state models. Limiting maximum system performance loss, a minimum quantification error aimed hierarchical clustering algorithm is designed to choose the most representative models. The system performance is improved by about 1.5% while saving 40% of the system expense. About 92% of the performance may also be maintained while reducing 70% of system parameters. The suggested method is quite useful for designing pen based interface for various handheld devices.
基金This work was supported by the Projects of the National Nat-ura! Science Foundation of China under Crant No.U0835001 the Fundamental Research Funds for the Central Universities-2011PTB-00-28.
文摘Abstract： A hierarchical method for scene analysis in audio sensor networks is proposed. This meth-od consists of two stages： element detection stage and audio scene analysis stage. In the former stage, the basic audio elements are modeled by the HMM models and trained by enough samples off-line, and we adaptively add or remove basic ele- ment from the targeted element pool according to the time, place and other environment parameters. In the latter stage, a data fusion algorithm is used to combine the sensory information of the same ar-ea, and then, a role-based method is employed to analyze the audio scene based on the fused data. We conduct some experiments to evaluate the per-formance of the proposed method that about 70% audio scenes can be detected correctly by this method. The experiment evaluations demonstrate that our method can achieve satisfactory results.
