Adaptive frequency cepstral coefficients for word mispronunciation detection

Systems based on automatic speech recognition (ASR) technology can provide important functionality in computer assisted language learning applications. This is a young but growing area of research motivated by the large number of students studying foreign languages. Here we propose a Hidden Markov Model (HMM)-based method to detect mispronunciations. Exploiting the specific dialog scripting employed in language learning software, HMMs are trained for different pronunciations. New adaptive features have been developed and obtained through an adaptive warping of the frequency scale prior to computing the cepstral coefficients. The optimization criterion used for the warping function is to maximize separation of two major groups of pronunciations (native and non-native) in terms of classification rate. Experimental results show that the adaptive frequency scale yields a better coefficient representation leading to higher classification rates in comparison with conventional HMMs using Mel-frequency cepstral coefficients.

I. Introduction

In this paper, we consider the problem of pronunciation detection and evaluation. This is an important challenge for which DSP algorithms can provide cost-effective solutions. Millions of people study foreign languages. Some are fortunate to have one-on-one time with the instructor, where they receive correction on words they mispronounce. Most students, however, do not. Since the number of language students out number instructors by factors of 20, 30, or more, instructors cannot effectively spend significant amounts of time with students individually.