Visiting Lecturer Program (241)

Published at: 2016-04-18

Shahram Kalantari - Copy.jpg
Speaker: Dr. Shahram Kalantari
PhD
Queensland University of Technology

Title: Audio visual speech recognition using hidden Markov models
Local Organizer: Dr. Hamidreza Zarandi
Time: Wednesday, April 20, 2016, 12:00- 13:30
Location: Department of Computer Engineering, Amirkabir University of Technology

Abstract:
Open vocabulary speech recognition requires training of speech models for different speech units. Due to limited number of publicly available and annotated speech databases, it is hard to obtain generalized models using current training methods. Acoustic noise is also an issue when using audio data for speech recognition. Visual data in the form of lip movements of the speaker has shown to improve the performance of speech recognition through audio visual model training methods. Multistream synchronous hidden Markov models have shown promising results for speech recognition through fused adaption of acoustic models on visual data.