Introduction: In recent years a number of methods based on acoustic analysis were developed for vocal fold pathology detection. These methods can be categorized in two categories: a) detection based on the phonemes b) detection based on the continuous speeches. While there are many researches which belong to the first category, there are few efforts for detecting vocal fold pathology based on the continuous speeches (second category). Methods: In this work, a method based on the Hidden Markov model Toolkit (HTK) for detecting vocal fold pathology in the Russian digits is developed which belongs to the second category. It employs a three state HMM for modeling each phoneme. Results: According to the results of the experiments, the proposed method achieves the 90% of detection accuracy. Conclusion: The proposed method is one of the first works for detecting vocal fold pathology based on the Russian digits (from 1 to 10) for Belorussian people. The reported accuracy is rather good and therefore it is recommended to use it as an auxiliary tool in medical centers.
Key words: vocal fold pathology, Automatic Speech Recognition (ASR), Hidden Markov model Toolkit (HTK), Russian digits.
|
