In this paper, we propose a speech recognition algorithm which utilizes hid
den Markov models (HMM) and Viterbi algorithm for segmenting the input spee
ch sequence, such that the variable-dimensional speech signal is converted
into a fixed-dimensional speech signal, called TN vector. We then use the f
uzzy perceptron to generate hyperplanes which separate patterns of each cla
ss from the others. The proposed speech recognition algorithm is easy for s
peaker adaptation when the idea of "supporting pattern" is used. The suppor
ting patterns are those patterns closest to the hyperplane. When a recognit
ion error occurs, we include all the TN vectors of the input speech sequenc
e with respect to the segmentations of all HMM models as the supporting pat
terns. The supporting patterns are then used by the fuzzy perceptron to tun
e the hyperplane that can cause correct recognition, and also tune the hype
rplane that resulted in wrong recognition. Since only two hyperplanes need
to be tuned for a recognition error, the proposed adaptation scheme is time
-economic and suitable for on-line adaptation. Although the adaptation sche
me cannot ensure to correct the wrong recognition right after adaptation, t
he hyperplanes are tuned in the direction for correct recognition iterative
ly and the speed of adaptation can be adjusted by a "belief" parameter set
by the user. Several examples are used to show the performance of the propo
sed speech recognition algorithm and the speaker adaptation scheme.