The Relationship Between Acoustic Signal Typing and Perceptual Evaluation of Tracheoesophageal Voice Quality for Sustained Vowels

Objectives
To investigate the relationship between acoustic signal typing and perceptual evaluation of sustained vowels produced by tracheoesophageal (TE) speakers and the use of signal typing in the clinical setting.

Methods
Two evaluators independently categorized 1.75-second segments of narrow-band spectrograms according to acoustic signal typing and independently evaluated the recording of the same segments on a visual analog scale according to overall perceptual acoustic voice quality. The relationship between acoustic signal typing and overall voice quality (as a continuous scale and as a four-point ordinal scale) was investigated and the proportion of inter-rater agreement as well as the reliability between the two measures is reported.

Results
The agreement between signal type (I-IV) and ordinal voice quality (four-point scale) was low but significant, and there was a significant linear relationship between the variables. Signal type correctly predicted less than half of the voice quality data. There was a significant main effect of signal type on continuous voice quality scores with significant differences in median quality scores between signal types I-IV, I-III, and I-II.

Conclusions
Signal typing can be used as an adjunct to perceptual and acoustic evaluation of the same stimuli for TE speech as part of a multidimensional evaluation protocol. Signal typing in its current form provides limited predictive information on voice quality, and there is significant overlap between signal types II and III and perceptual categories. Future work should consider whether the current four signal types could be refined.

Key Words
Automatic evaluation; Head and neck cancer; Perceptual evaluation; Acoustic signal typing; Tracheoesophageal speech; Laryngectomy