Algorithm and DSP-implementation for a speaker-independent single-word speech recognizer with additional speaker-dependent say-in facility

We present a novel system implementation concept for small-vocabulary single-word speech recognition. The system can recognize speaker-independently a number of preprogrammed words, and additionally it can perform recognition of a number of words which have been trained by simple say-in. In contrast to other implementations, one and the same algorithm is used for both speaker-dependent and speaker-independent recognition. This implies that in our proposed system the fixed (speaker-independent) part and the personalized (speaker-dependent) part of the vocabulary can be activated simultaneously, which is mandatory for the use in mobile communication systems. Compared to state-of-the-art speaker-dependent recognition systems, the RAM capacity for storing the personalized part of the vocabulary is drastically reduced. The recognition rate is comparable to that of a fully speaker-independent system. A DSP implementation of the speech recognition system is described.