Improving Robustness in Jacobian Adaptation for Noisy Speech Recognition

  • Authors:

  • Yongjoo Jung

  • Affiliations:

  • Department of Electronics, Keimyung University, Daegu, S. Korea

  • Venue:
  • PIT '08 Proceedings of the 4th IEEE tutorial and research workshop on Perception and Interactive Technologies for Speech-Based Systems: Perception in Multimodal Dialogue Systems
  • Year:
  • 2008

Quantified Score

Hi-index 0.00

Visualization

Abstract

A method to improve the robustness of the Jacobian adaptation (JA) is proposed. Although it is a usual idea that the reference hidden Markov model (HMM) in the JA is constructed by using the model composition methods like the parallel model combination (PMC), we propose to train the reference HMM directly with the noisy speech and then select the appropriate reference HMM based on the noise types and signal to noise ratio (SNR) values obtained from the input noisy speech. For the estimation of Jacobian matrices and other statistical information for the JA, a data driven method is employed during the training.