An improved uncertainty decoding scheme with weighted samples for DNN-HMM hybrid systems

2016/08/04 by Huemmer, Christian, Astudillo, Ramón Fernández, Kellermann, Walter
#Computation and Language (cs.CL) #FOS: Computer and information sciences #Machine Learning (cs.LG) #Sound (cs.SD)

paper · doi:10.48550/arxiv.1609.02082

Abstract

In this paper, we advance a recently-proposed uncertainty decoding scheme for DNN-HMM (deep neural network - hidden Markov model) hybrid systems. This numerical sampling concept averages DNN outputs produced by a finite set of feature samples (drawn from a probabilistic distortion model) to approximate the posterior likelihoods of the context-dependent HMM states. As main innovation, we propose a weighted DNN-output averaging based on a minimum classification error criterion and apply it to a probabilistic distortion model for spatial diffuseness features. The experimental evaluation is performed on the 8-channel REVERB Challenge task using a DNN-HMM hybrid system with multichannel front-end signal enhancement. We show that the recognition accuracy of the DNN-HMM hybrid system improves by incorporating uncertainty decoding based on random sampling and that the proposed weighted DNN-output averaging further reduces the word error rate scores.

Related

• Uncertainty weighting and propagation in DNN–HMM-based speech recognition
• DNN-based uncertainty estimation for weighted DNN-HMM ASR
• Information Theoretic Analysis of DNN-HMM Acoustic Modeling
• Optimize what matters: Training DNN-HMM Keyword Spotting Model Using End\n Metric
• Increasing the Interpretability of Recurrent Neural Networks Using Hidden Markov Models