Fuzzy k-nearest neighbors applied to phoneme recognition

Abstract

In this work, the Fuzzy kNN (FkNN), an alternative of the standard kNN algorithm, is used for Timit phoneme recognition. Phoneme is the smallest unit that composes speech. For this reason, if phoneme recognition is performed, it can achieve a significant word and text recognition. Thus, the main idea consists on assigning phoneme membership to the data phonemes by measuring the distance to its kNN. FkNN compute the fuzzy distances between the data phonemes that define the cluster fuzziness. Mel Frequency Cepstral Coefficients (MFCC) associated with their first and second derivatives and energy coefficient were extracted from the speech signals as an input of the recognition system. A comparison of a crisp and fuzzy kNN was performed. Experiments show that FkNN algorithm not only can lead to significant recognition rates, but also may supersede in some ways Hidden Markov Model (HMM) the reference of speech recognition.