Deterministic annealing EM algorithm in parameter estimation for acoustic model
نویسندگان
چکیده
This paper investigates the effectiveness of the DAEM (Deterministic Annealing EM) algorithm in acoustic modeling for speaker and speech recognition. Although the EM algorithm has been widely used to approximate the ML estimates, it has the problem of initialization dependence. To relax this problem, the DAEM algorithm has been proposed and confirmed the effectiveness in small tasks. In this paper, we applied the DAEM algorithm to speaker recognition based on GMMs and continuous speech recognition based on HMMs. Experimental results show that the DAEM algorithm can improve the recognition performance as compared to the ordinary EM algorithm with conventional initialization methods, especially in the flat start training for continuous speech recognition.
منابع مشابه
Deterministic Annealing Variant of the EM Algorithm
We present a deterministic annealing variant of the EM algorithm for maximum likelihood parameter estimation problems. In our approach, the EM process is reformulated as the problem of minimizing the thermodynamic free energy by using the principle of maximum entropy and statistical mechanics analogy. Unlike simulated annealing approaches, this minimization is deterministically performed. Moreo...
متن کاملNBCT: A Toolkit for Naive Bayes Clustering
3 Clustering algorithms 5 3.1 ML/MAP based clustering . . . . . . . . . . . . . . 5 3.1.1 Parameter estimation based on ML . . . . . . 5 3.1.2 Parameter estimation based on MAP . . . . . 5 3.1.3 Membership distribution . . . . . . . . . . . 6 3.1.4 Dissimilarity . . . . . . . . . . . . . . . . . 7 3.1.5 Clustering . . . . . . . . . . . . . . . . . . . 7 3.1.6 Relevance analysis . . . . . . . . ...
متن کاملAn average case performance of the deterministic annealing EM algorithm
Average case performance of the deterministic annealing EM algorithm is evaluated for Gaussian mixture estimation problem under some additive noises. The data-averaged EM update equations with respect to hyperparameters are calculated analytically in the large data limit. We find that the EM algorithm strongly depends on the initial conditions. Moreover, by using our analysis, it becomes possib...
متن کاملHarmonic-temporal-structured Clustering via Deterministic Annealing Em Algorithm for Audio Feature Extraction
This paper proposes “harmonic-temporal structured clustering (HTC) method”, that allows simultaneous estimation of pitch, intensity, onset, duration, etc., of each underlying source in multi-stream audio signal, which we expect to be an effective feature extraction for MIR systems. STC decomposes the energy patterns diffused in timefrequency space, i.e., a time series of power spectrum, into di...
متن کاملDeterministic Annealing EM Algorithm in Acoustic Modeling for Speaker and Speech Recognition
This paper investigates the effectiveness of the DAEM (Deterministic Annealing EM) algorithm in acoustic modeling for speaker and speech recognition. Although the EM algorithm has been widely used to approximate the ML estimates, it has the problem of initialization dependence. To relax this problem, the DAEM algorithm has been proposed and confirmed the effectiveness in artificial small tasks....
متن کامل