Perceptual evaluation for automatic anomaly detection in disordered speech: Focus on ambiguous cases

Abstract

Perceptual evaluation is still the most common method in clinical practice for diagnosing and following the condition progression of people suffering from dysarthria (or speech disorders more generally). Such evaluations are frequently described as non-trivial, subjective and highly time-consuming (depending on the evaluation level). Most of the time, perceptual assessment is performed individually by clinicians which can be problematic since judgment may vary from one clinician to the other. Clinicians have therefore expressed the need for new objective evaluation tools better adapted to longitudinal studies, the observation of small units and rehabilitation context to monitor patients' progress. We have previously proposed an automatic approach to the anomaly detection at the phone level for dysarthric speech. The system behavior was studied and validated with different corpora and speech styles and shows good results in this specific task. Nonetheless, the lack of annotated French dysarthric speech corpora has limited our ability to analyze some aspects of its behavior, such as severity, more precisely (more anomalies are detected automatically compared with human experts). To overcome this limitation, we proposed an original perceptual evaluation protocol applied to a limited set of decisions made by the automatic system, relating to the presence of anomalies. Particularly, we intended to focus our analyses on some ambiguous cases in order to enrich our knowledge about the

Construction of an automatic score for the evaluation of speech disorders among patients treated for a cancer of the oral cavity or the oropharynx: The Carcinologic Speech Severity Index

“…Different studies involving automatic speech processing had been conducted on a first batch of recordings before being applied to the whole corpus. Among these studies, we can report the use of an i‐vectors‐based approach derived from the automatic speaker recognition field (automatically recognizes the identity of a speaker based on his/her voice) for predicting a speech intelligibility score 41 ; the use of an automatic system for detecting speech abnormalities, especially designed for analyzing speech impairments 42 ; and the proposal of a “robotic” listener for automatically performing the phonetic‐acoustic decoding of the pseudoword productions and for providing an automatic measure of speech intelligibility 43 . Automatic speech alignment was involved in the last two studies, which consists of aligning the sequence of expected phonemes (corresponding to a word or a pseudoword pronounced by a speaker in this case) on the corresponding speech signal.…”

Principes et outils pour l’annotation des corpus

“…C'est le cas pour les différentes maladies caractérisées par une dysarthrie [voir l'article Phonétique Clinique dans ce numéro] impactant la motricité des gestes nécessaires à la production des sons de la parole. Si une attention particulière a été donnée depuis de nombreuses années au développement d'outils automatiques adaptés au traitement automatique de la parole pathologique (Fredouille & Pouchoulin, 2012 ;Laaridh et al, 2018), il n'en reste pas moins que l'AA de ce type de parole doit parfois être corrigé. Dans ce cas, les critères classiques de segmentation phonétique doivent être adaptés et il est nécessaire d'intégrer des classes de segments spécifiques à ce type de parole (voir Meunier, 2014, pour un développement de ces aspects).…”

30 ans de recherches en phonétique clinique au LPL

“…La contribution du Laboratoire Informatique d'Avignon (C. Fredouille, LIA) et la thèse d'I. Laaridh (financée du Brain & Language Research Institute, BLRI) ont permis des avancées considérables concernant l'identification automatique des zones déviantes dans la parole dysarthrique(Laaridh et al, 2016 ;Laaridh et al, 2018) sans laquelle les analyses phonétiques de ces déviances auraient été rendues très difficiles.…”