Enhancement of esophageal speech using formant synthesis

The feasibility of using the formant analysis-synthesis approach to replace the voicing sources of esophageal speech was explored. The voicing sources were generated by using inverse-filtered signals extracted from normal speakers. Pitch extraction was tested with various pitch extraction methods, then a simple auto-correlation method was chosen. A special hardware unit was designed to perform the analysis-synthesis process in real-time. Results of a subjective test showed that the synthesized speech was significantly improved