Effective Glottal Instant Detection and Electroglottographic Parameter Extraction for Automated Voice Pathology Assessment.

Abstract

Accurate determination of glottal instants and electroglottographic (EGG) parameters is most important in voice pathology analysis including multiple voice disorders: neurological, functional, and laryngeal diseases. In this paper, we present a new effective method for reliable detection of glottal instants and EGG parameters from an EGG signal composed of voiced and nonvoice segments. In the first stage, we present an adaptive variational mode decomposition based algorithm for suppressing low-frequency artifacts and additive high-frequency noises. Based upon mode center frequency criterion, the proposed method first constructs a candidate EGG feature signal for determination of glottal closure and opening instants. In the second stage, the candidate glottal instants are determined by detecting the positive and negative zerocrossings in normalized candidate EGG feature signal, respectively. Finally, an autocorrelation features based postprocessing algorithm is presented to reject nonglottal instants from the nonspeech production segments. The accuracy and robustness of the method is tested using noise-free and noisy EGG signals. Evaluation results show that the proposed method achieves an average overall accuracy of 95.06%, identification rate of 95.34%, missed rate of 3.60%, and false alarm rate of 0.06% with average absolute identification error of 0.71 ± 0.66ms for an SNR of 15dB. Results demonstrate that the proposed method significantly outperforms the other existing methods under both noise-free and noisy EGG signals.