Analysis of voice features of people with cervical spinal cord injuries with voice user interfaces

Abstract

Abstract Study design: A comparative study Objectives: To evaluate respiratory function, voice intensity, self-assessment of voice problems, and recordings in individuals with cervical spinal cord injuries (CSCIs), compare these parameters with a control group, and identify factors influencing speech intelligibility. Setting: Clinical setting at the Spinal Injuries Center Methods: Twenty individuals with CSCIs and 20 healthy participants were recruited for this study. Their respiratory function was assessed using an electric diagnostic spirometer, and voice intensity and recordings were measured using a smartphone sound analysis application. Self-assessed voice problems were evaluated using the Voice Handicap Index. Statistical analyses were conducted to compare the measurement data between the two groups. A simple regression analysis was used to assess the correlation between speech intelligibility and various vocal parameters. Results: Vital capacity was 60% lower in the CSCI group. Maximum voice intensity in the CSCI group (mean 78.8 dB) was significantly lower than that in the control group (mean 96.2 dB) (P<0.001). In the CSCI group, significant correlations between speech intelligibility and maximum phonation time (R2=0.298, P=0.013), maximum vocalization voice intensity (R2=0.673, P<0.001), reading voice intensity (R2=0.664, P<0.001), and the smoothed cepstral peak prominence value (R2=0.437, P=0.002) were observed. A weak correlation between vital capacity and speech intelligibility was also noted. Conclusions: Clinically, individuals with CSCIs were not consciously aware of their voice problems, despite exhibiting low speech intelligibility in voice user interfaces. Voice recognition performance could be improved through voice conversion techniques that increase voice intensity and speech rate. Sponsorship: not applicable