On Naive Bayes In Speech Recognition - Skripsi Teknik Informatika

On Naive Bayes In Speech Recognition

(Jurnal Teknik Informatika)

ABSTRAK

The currently dominant speech recognition technology, hidden Markov modeling, has long been criticized for its simplistic assumptions about speech, and especially for the naive Bayes combination rule inherent in it. Many sophisticated alternative models have been suggested over the last decade. These, however, have demonstrated only modest improvements and brought no paradigm shift in technology. The goal of this paper is to examine why HMM performs so well in spite of its incorrect bias due to the naive Bayes assumption. To do this we create an algorithmic framework that allows us to experiment with alternative combination schemes and helps us understand the factors that influence recognition performance. From the findings we argue that the bias peculiar to the naive Bayes rule is not really detrimental to phoneme classification performance. Furthermore, it ensures consistent behavior in outlier modeling, allowing efficient management of insertion and deletion errors.

Keywords: naive Bayes, segment-based speech recognition, hidden Markov model

1. Introduction

Although speech recognition requires the fusion of several information sources, it is rarely viewed as an expert combination problem. Such approaches were abandoned in favor of the hidden Markov modeling technique (HMM) (Huang et al., 2001), which treats speech as a stochastic process. The source of the success of HMM is that it offers a sound mathematical framework along with efficient training and evaluation. The price is that the simplistic mathematical assumptions of the model do not accord with the real behavior of speech. One of these assumptions is the conditional independence of the spectral vectors. Several alternative models have been proposed to alleviate this flaw, but these have brought only modest improvements at the cost of a considerable increase in complexity. Rather than seeking to eliminate the incorrect modeling bias, here we hope to gain a better insight into whyHMM performs so well in spite of the unrealistic naive Bayes assumption.

Untuk lebih lengkapnya anda bisa mengdownload jurnal nya di link berikut :

Kata Kunci : Jurnal International, Jurnal Teknik Informatika, Jurnal Skripsi, Jurnal, Contoh Jurnal, Skripsi Teknik Informatika,Contoh Skripsi, Skripsi.