1. Home
    Item seperator
  3. Item seperator
    A Probabilistic Acoustic Map Based Discriminative HMM Training

A Probabilistic Acoustic Map Based Discriminative HMM Training

New Image

In this paper, based upon a probabilistic acoustic map (PAM) framework, we propose a discriminative Hidden Markov Model (HMM) training procedure. The goal of this new training procedure is to improve the discriminative power of a maximum likelihood (ML) HMM without sacrificing its classification capability.