Comparison of HMM experts with MLP experts in the Full Combination Multi-Band Approach to Robust ASR

摘要: In this paper we apply the Full Combination (FC) multi-band approach, which has originally been introduced in framework of posterior-based HMM/ANN (Hidden Markov Model/Artificial Neural Network) hybrid systems, to systems ANN (or Multilayer Perceptron (MLP)) is itself replaced by a Multi Gaussian HMM (MGM). Both represent most widely used statistical models for robust ASR (automatic speech recognition). It shown how FC formula likelihood-based MGMs can easily be derived from approach simply applying Bayes' Rule. The experiments show that system with MGM experts performs better, all noise conditions tested, than simple sum and product rules are normally used. As compared baseline full-band system, shows increased robustness mainly on band-limited noise. goal article not performance comparison between Perceptrons Models but theory two approaches, vs. so results only given MGMs.