Gaussian Mixture Model Based Classification of Stuttering Dysfluencies

Abstract

The classification of dysfluencies is one of the important steps in objective measurement of stuttering disorder. In this work, the focus is on investigating the applicability of automatic speaker recognition (ASR) method for stuttering dysfluency recognition. The system designed for this particular task relies on the Gaussian mixture model (GMM), which is the most widely used probabilistic modeling technique in ASR. The GMM parameters are estimated from Mel frequency cepstral coefficients (MFCCs). This statistical speakermodeling technique represents the fundamental characteristic sounds of speech signal. Using this model, we build a dysfluency recognizer that is capable of recognizing dysfluencies irrespective of a person as well as what is being said. The performance of the system is evaluated for different types of dysfluencies such as syllable repetition, word repetition, prolongation, and interjection using speech samples from the University College London Archive of Stuttered Speech (UCLASS).

An Automatic Prolongation Detection Approach in Continuous Speech With Robustness Against Speaking Rate Variations

“…In this study, 39 audio recordings of individuals (i.e., 37 males and 2 females) with age ranging from 8 years and 4 months to 20 years and 1 month were taken from UCLASS database. As suggested by other researchers,[81012] the samples were taken from “reading” category that mostly contains monosyllabic words. Because most of the recordings do not have labels, two well-trained SLPs were employed to label the 39 samples of our choice into fluent or prolongation classes.…”

“…[5] To support objective classification of speech dysfluencies, several researches have been performed on the classification of prolongation either individually[67] or in combination with other dysfluencies. [8910]…”

“…Then, different sets of features and classifiers were used to detect prolongations among normal samples. Mel frequency cepstral coefficient (MFCC) and hidden Markov model (HMM),[6] entropy of wavelet packet transforms (WPT) and support vector machine (SVM),[8] MFCC and Gaussian mixture model (GMM),[10] speech envelop and artificial neural network (ANN),[11] and linear predictive cepstral coefficient (LPCC) and K-nearest neighbor (K-NN)[12] are a few examples of such procedures.…”

“…The segments were then labeled as prolongation if the duration of highly similar successive frames exceeded a threshold specified by the speaking rate. All previous studies have employed either a fixed threshold for minimum sustaining duration in sounds[67] or classifiers that were trained on predefined speaking rates,[891012] although it must actually be judged according to how slow or fast the individual’s speech is. The performance of the proposed method was appraised using English and Persian language databases by comparing our results with those of three accurate methods[789] from the previous researches.…”

SEP-28k: A Dataset for Stuttering Event Detection from Podcasts with People Who Stutter

“…The focus of this paper is on detection of five stuttering event types: Blocks, Prolongations, Sound Repetitions, Word/Phrase Repetitions, and Interjections. Existing work has explored this problem using traditional signal processing techniques [15,16,17], language modeling (LM) [12,18,19,20,21], and acoustic modeling (AM) [21,10]. Each approach has be shown to be effective 1.…”