Form hybrid adjustable neuro-fuzzy systems to adaptive connectionist-based systems for phoneme and word recognition

摘要

This paper discusses the problem of adaptation in automatic speech recognition systems (ASRS) and suggest several strategies for adaptation in a modular architecture for speech recognition. The architecture allow for adaptation at different levels of the recognition process, where modules can be adapted individually based on their performance and the performance of the whole system. Two realizations of this architecture are presented along with experimental results form small-scale experiments. The first realization is a hybrid system for spear-independent phoneme-based spoken word recognition, consisting of neural networks for recognizing English phonemes and fuzzy systems for modeling acoustic and linguistic knowledge. This system is adjustable by addition training of individual neural network modules and tuning the fuzzy systems. The increase accuracy of the recognition through appropriate adjustment is also discussed. The second realization of the architecture is a connectionist system that uses fuzzy neural networks FuNNs to accommodate both a prior linguistic knowledge and data from a speech corpus. A method for on-line adaptation of FuNNs is also presented.