Machine Learning for Stuttering Identification: Review, Challenges and Future Directions

Abstract

Stuttering is a speech disorder during which the flow of speech is interrupted by involuntary pauses and repetition of sounds. Stuttering identification is an interesting interdisciplinary domain research problem which involves pathology, psychology, acoustics, and signal processing that makes it hard and complicated to detect. Recent developments in machine and deep learning have dramatically revolutionized speech domain, however minimal attention has been given to stuttering identification. This work fills the gap by trying to bring researchers together from interdisciplinary fields. In this paper, we review comprehensively acoustic features, statistical and deep learning based stuttering/disfluency classification methods. We also present several challenges and possible future directions.

Robust Stuttering Detection via Multi-task and Adversarial Learning

“…Speech impairments are speech disorders in which a speaker toils to create the speech sounds required for communication [1]. These speech impairments can take different forms which include apraxia, cluttering, dysarthria, stuttering, etc [2]. Among the speech disorders, stuttering -also known by the name stammering/disfluency -has been found to be the most common one [2].…”

“…These speech impairments can take different forms which include apraxia, cluttering, dysarthria, stuttering, etc [2]. Among the speech disorders, stuttering -also known by the name stammering/disfluency -has been found to be the most common one [2]. Stuttering is a neuro-developmental speech impairment which occurs due to the malfunctioning of sensorimotors, responsible for speech production [3], and usually take the shape of core behaviours: blocks, repetitions and prolongations [1].…”

“…The advancements of deep learning (DL) has shown immense development in various speech domains like speech synthesis [5], ASR [6], emotion detection [7], speaker identification [8], etc, however SI/SD has received minimal attention from the DL paradigm. DL can help in detecting the various types of stuttering by exploiting the presence of various acoustic cues in stuttered speech [2]. The recent works have made significant strides in DL based SD [2], however, all the existing systems focus only on single task stuttering learning strategy without focusing on the auxiliary tasks.…”

“…DL can help in detecting the various types of stuttering by exploiting the presence of various acoustic cues in stuttered speech [2]. The recent works have made significant strides in DL based SD [2], however, all the existing systems focus only on single task stuttering learning strategy without focusing on the auxiliary tasks. Unfortunately, this is not the approach that we human beings process the speech utterances, rather we simultaneously decode the primary task with other meta contents like speaker characteristics, linguistic content, etc.…”

“…SD systems are further affected by source diversity, be that linguistic content, speaker, gender, accent, or encompassing acoustic conditions [2]. In this study, in addition to MTL, we also investigate an ADV framework with the aim to learn robust stuttering representations.…”