Multistage Switched Adaptive Filtering Approach for Denoising Speech Signals of Parkinson's Disease-affected Patients

Recording the speech signals of Parkinson's Disease (PD)-affected patients is challenging due to the surrounding noise. Therefore there is a need to denoise the signals. This paper proposes an Adaptive Noise Canceller-based model for signal denoising. This paper introduces an optimal adaptive filter structure using a signed LMS algorithm to compute the best estimate of a clean signal. A noise-corrupted signal is sent across multiple adaptive filters connected in series. Multiple stages are added automatically, and the filtering algorithm for each stage is also adjusted automatically. The proposed multi-stage switched adaptive filter model is tested for reducing the noise from a speech signal recorded from Parkinson's Disease-affected patients and corrupted by Gaussian signals of different input SNR levels. The simulation results prove that the proposed filter model performs remarkably well and provides 20-30 dB higher SNR values than the existing cascaded LMS filter models. The MSE value is improved by 85-97%, and the PSNR values are increased by 7 dB. Using the Sign LMS algorithm in the proposed filter model offers a cost-effective hardware implementation of Adaptive Noise Canceller with high accuracy.

Automated summary

This paper proposes an Adaptive Noise Canceller-based model for denoising speech signals recorded from Parkinson's Disease-affected patients. The model uses a signed LMS algorithm to compute the best estimate of a clean signal, and consists of multiple adaptive filters connected in series to reduce noise. The simulation results show that the proposed filter model outperforms existing cascaded LMS filter models in terms of signal-to-noise ratio improvement.