Advances in automatic speech recognition by combining information theory and neural network algorithms
Gerhard Rigoll · International Journal of Quantum Chemistry · 1992
Abstract This paper presents some of the newest developments in the field of automatic speech recognition. It concentrates on describing the efforts of combining the two so‐far most successful approaches in automatic speech recognition, namely, information theory and neural network techniques, in order to obtain more improved speech recognition algorithms. An overview of the various approaches to the combination of these two basic technologies is given and the arising problems, such as the choice of the appropriate neural network paradigms, the best possible combination strategies and possibilities for an integrated training of the neural network, and the Markov model parameters, are discussed.