An on-line automatic spoken word recognition system has been developed for the researches on the automatic recognition of speech. In this system, the spoken word is, first, converted into a time series of short time spectra by 29-channel filter bank of single tuned low selectivity filters. Three major local peaks in the spectrum and the power of the speech wave are extracted in every 10 ms. Input speech is transformed into some possible phonemic sequences by using three major local peaks and the speech power. The similarity of the sequence to every item of the word dictionary in the recognition system is computed. The item of the dictionary having the maximum similarity to the sequence is chosen as the output of the recognition. Some recognition experiments have been carried out with the system. In the interactive experiment, the recognition score was found to be 94% for 51 city names uttered by 25 male speakers arbitrarily chosen.

To investigate necessary conditions for the object recognition by simulations using neural network models is one of ways to acquire suggestions for understanding the neuronal representation of objects in the brain. In the present study, we trained a three layered neural network to form a geometrical feature representation in its output layer using back-propagation algorithm. After training using 73 learning examples, 65 testing patterns made by various combinations of above features could be recognized with the network at a rate of 95.3% appropriate response. We could classify four types of hidden layer units on the basis of effects on the output layer.

We reported that a competitive learning neural network had the ability of self-organization in the classification of questionnaire survey data. In this letter, its self-organized learning was evaluated by means of mutual information. Mutual information may be useful to find efficently the network which can give optimal classification.