Continuous Speech Recognition Using Modified Stack Decoding Algorithm

This paper attempts to recognize a speech sequence with a series of words. We represent words with HMM models and use a modified version of stack decoding method as the primary algorithm to recognize words. This type of method allows us to recognize a streaming sequence of speech signal as we receive it. We introduced a heuristic function that works with stack decoding algorithm. We also observed how model parameters affect the performance of the algorithm.