An improved automatic lipreading system to enhance speech recognition

摘要: Current acoustic speech recognition technology performs well with very small vocabularies in noise or large low noise. Accurate over 100 words has yet to be achieved. Humans frequently lipread the visible facial articulations enhance recognition, especially when signal is degraded by hearing impairment. Automatic lipreading been found improve significantly and could advantageous noisy environments such as offices, aircraft factories.An improved version of a previously described automatic system developed which uses vector quantization, dynamic time warping, new heuristic distance measure. This paper presents visual results from multiple speakers under optimal conditions. Results combined are also presented show performance compared alone.