Cochlea-spectral entropy and intelligibility in spastic dysarthria associated with cerebral palsy

Abstract

The cochlea-scaled entropy (CSE) measure has been suggested as a measure of speech intelligibility in healthy speakers and speakers with hypokinetic dysarthria associated with Parkinson's disease (PD). However, the reliability of CSE measure use in speakers with spastic dysarthria associated with cerebral palsy (CP) is unknown. The aims of this study were to discover whether the CSE measure is a reliable measure for assessing speech intelligibility in people with spastic dysarthria associated with CP and whether the CSE measure is able to differentiate between healthy speakers and speakers with spastic dysarthria associated with CP. The performances of CSE parameters in the study are also discussed. Data from 25 speakers (12 in the CP group and 13 in the control group) was taken from the UA-Speech Database (Kim et al. 2008). The CP speakers were divided into groups according to their level of speech intelligibility: levels 4 (high), 3 (medium), 2 (low) and 1 (very low). Twenty-four single-word speech samples with three recording repetitions were analysed and were then divided into easy phoneme and difficult phoneme group for the investigation of the CSE performance between the groups. A Pearson correlation coefficient test was carried out to examine the correlation between the CSE measure and speech intelligibility in the CP group. Boxplots and t-tests were carried out to investigate the performance of the CSE values between the control and the CP groups as well as the CP subgroups. The CSE performances in the easy and difficult phoneme groups between the control and the CP groups were also examined using t-tests. The correlation between the CSE measure and the speech intelligibility in the CP group was not significant (p>.05). The CSE values between the control group and the CP group were significant different (p<.05) and the CSE values were higher in the control group (mean=13.01; SD=4.73; minimum=5.21 and maximum=48.12) than in the CP group (mean=12.42; SD=3.92; minimum=3.99 and maximum=41.53) as well as than other CP subgroups. The CSE standard deviation in the 12 words in the easy phoneme group and 10 out of 12 words in the difficult phoneme group showed significant differences (p<.05) between the control and the CP group. These findings suggest that CSE does not correlate with ratings of speech intelligibility in the CP group but the CSE measure was able to differentiate the control and the CP group. Moreover, the CSE SD may be the most sensitive parameter to differentiate healthy speakers and speakers with spastic dysarthria associated with CP when single word speech samples contain easy phonemes and one syllable.