[Research on Diagnosis Algorithm of Parkinson’s Disease Based on Speech Sample Multi-edit and Random Forest]

Area of Science:

• Biomedical Engineering
• Computational Linguistics
• Neurology

Background:

• Speech analysis is an emerging, effective method for Parkinson's disease (PD) diagnosis.
• Existing methods face challenges in sample preprocessing and ensemble learning, leading to suboptimal classification accuracy and stability.

Purpose of the Study:

• To develop a novel algorithm for accurate and stable diagnosis of Parkinson's disease using speech data.
• To address limitations in current speech-based PD diagnostic approaches.

Main Methods:

• A new algorithm combining multi-edit sample selection with random forest was proposed.
• Experiments were conducted using the latest public speech datasets for PD diagnosis.

Main Results:

• The algorithm successfully classified both PD samples and subjects.
• Achieved an average classification accuracy of 100%, an improvement of up to 29.44% over existing methods.
• Demonstrated higher and more stable classification accuracy compared to other algorithms.

Conclusions:

• The proposed speech diagnosis algorithm for PD, based on instance selection, offers superior performance.
• This method provides a more accurate and stable approach for identifying Parkinson's disease from speech patterns.