A novel low complexity speech recognition approach

In the field of digital speech recognition powerful ASR (automatic speech recognizer) systems have been developed which employ highly intricate algorithms like the HMM, DTW and neural network based algorithms capable of recognizing up to 1000 different words. Their high complexity and computation requirements prove to be superfluous for less demanding tasks. In this paper is proposed a simple, less aggressive and computationally efficient algorithm that can parallel any of the above algorithmspsila ability to distinguish a few words (typically up to 5 different words) but demands considerably less computational power making it suitable for embedded systems. Also discussed is the technique of rdquoMenu Drivenrdquo control where the number of words that can be recognized, pose no frontier to the number of tasks that can be performed by using very few (5 or less) words.