Assessing dysarthria using variability measures from audio recordings

Classification and characterization of motor speech disorders (MSDs) is important from the viewpoint of diagnosis and treatment. Clinical diagnosis is primarily based on auditoryperceptual characteristics of perceived speech abnormalities, but is subject to unreliable clinical judgement and quantifi cation, and diffi cult to relate to the underlying pathophysiology. In this study we investigate whether it is possible to diagnose dysarthria based on measures of speech variability by using Functional Data Analysis (FDA) (Ramsay et al.,1996). A reliable quantifi cation of variability in speech can potentially reveal underlying motor control problems, enable early detection of sub-clinical speech abnormalities, and provide sensitive and quantifi able outcome measures that aid treatment strategies. FDA has
been shown to be successful in investigating variability of kinematic movements obtained by lip displacement tracking, but may also be applied to other dimensions of speech, including amplitude envelopes and pitch and formant tracks. Anderson et al. (2008) used FDA to calculate spatial and temporal variability of amplitude envelopes of sentence repetitions produced by patients with hypokinetic and ataxic dysarthria and demonstrated that variability
characteristics were infl uenced by dysarthria type.