Artificial intelligence empowered voice generation for amyotrophic lateral sclerosis patients

Area of Science:

• Neuroscience and Speech Technology
• Biomedical Engineering
• Artificial Intelligence in Healthcare

Background:

• Amyotrophic Lateral Sclerosis (ALS) progressively impairs speech, impacting patients' mental well-being.
• Alternative Augmentative Communication (AAC) using synthetic voices aids communication but often lacks personal identity.
• Voice banking (VB) combined with AI offers a personalized solution for preserving patient voices.

Purpose of the Study:

• To assess the effectiveness and perceptual quality of AI-generated personalized voices for ALS patients with preserved speech.
• To investigate the implementation of voice banking in rapidly progressing diseases like ALS.
• To understand the early-stage perceptual impact of VB on ALS patients' communication.

Main Methods:

• Utilizing a machine learning-based personalized voice synthesis system for ALS patients.
• Employing voice recording and fine-tuning with Generative Adversarial Networks (HiFi-GAN) for high-fidelity speech synthesis.
• Evaluating AI-generated patient-specific voices for exceptional expressive and audio quality.

Main Results:

• AI-generated voices closely mimic the patient's original voice, offering high expressive and audio quality.
• The study provides insights into the benefits and challenges of AI-driven voice banking for ALS.
• Personalized synthetic voices show potential for enhancing communication support in ALS.

Conclusions:

• AI-powered voice banking offers a promising approach to maintain patient identity and improve communication for individuals with ALS.
• Further research is needed to optimize VB strategies for rapidly progressing neurological conditions.
• Personalized synthetic voices can significantly enhance the quality of life for ALS patients by preserving their unique vocal identity.