PVR-Vocoder: A Pathological Voice Repair Vocoder for Voice Disorders

Area of Science:

• Speech synthesis and signal processing.
• Medical acoustics and voice pathology.

Background:

• Existing vocoder techniques primarily focus on synthesizing normal speech.
• Pathological voice disorders present unique challenges in speech synthesis and require specialized solutions.
• Current methods often neglect the specific needs of individuals with voice impairments, impacting speech intelligibility.

Purpose of the Study:

• To propose a non-linear voice repair vocoder (PVR-Vocoder) for pathological speech.
• To enhance the quality and intelligibility of pathological vowels and sentences.
• To address the limitations of existing vocoders in handling voice disorders.

Main Methods:

• Utilizing amplitude modulated-frequency modulated (AM-FM) and Teager energy operations for pitch and spectral envelope enhancement.
• Implementing a spectral tracking algorithm to stabilize pitch and reduce errors.
• Developing a spectral reconstruction algorithm for rebuilding spectral structure via energy operations.

Main Results:

• The PVR-Vocoder demonstrates exceptional performance in enhancing pathological voice intelligibility.
• Objective indicators, subjective evaluations, and spectrum observations confirm the vocoder's effectiveness.
• Significant improvements in speech quality and intelligibility for pathological voices were achieved.

Conclusions:

• The proposed PVR-Vocoder offers a viable solution for pathological voice repair.
• This technique has the potential to significantly improve speech interaction for individuals with voice disorders.
• The study highlights the importance of specialized vocoder approaches for addressing specific speech pathologies.