Affiliations: [a] I.K. Gujral Punjab Technical University (PTU), Jalandhar, Punjab-144001, India | [b] Department of Electronics and Communication, University Institute of Engineering and Technology (UIET), Panjab University (PU), Chandigarh-160023, India | [c] Department of Electronics and Communication, BBSBEC, Fatehgarh Sahib, Punjab-140407, India | [d] Central Scientific Instruments Organisation (CSIR-CSIO), Chandigarh-160030, India
Correspondence:
[*]
Corresponding author: Nidhi Garg, I.K. Gujral Punjab Technical University (PTU), Jalandhar, Punjab-144001, India and Department of Electronics and Communication, University Institute of Engineering and Technology (UIET), Panjab University (PU), Chandigarh-160023, India. Tel.: +918727058014; E-mails: nidhi_garg@pu.ac.in, nidhi461@gmail.com.
Abstract: Background: WPS is a non-invasive method to investigate human health. During signal acquisition, noises are also recorded along with WPS. Objective: Clean WPS with high peak signal to noise ratio is a prerequisite before use in disease diagnosis. Wavelet Transform is a commonly used method in the filtration process. Apart from its extensive use, the appropriate factors for wavelet denoising algorithm is not yet clear in WPS application. The presented work gives an effective approach to select various factors for wavelet denoise algorithm. With the appropriate selection of wavelet and factors, it is possible to reduce noise in WPS. Methods: In this work, all the factors of wavelet denoising are varied successively. Various evaluation parameters such as MSE, PSNR, PRD and Fit Coefficient are used to find out the performance of the wavelet denoised algorithm at every one step. Results: The results obtained from computerized WPS illustrates that the presented approach can successfully select the mother wavelet and other factors for wavelet denoise algorithm. The selection of db9 as mother wavelet with sure threshold function and single rescaling function using UWT has been a better option for our database. Conclusion: The empirical results proves that the methodology discussed here could be effective in denoising WPS of any morphological pattern.
Keywords: Wrist pulse signal, signal processing, wavelet denoising, PSNR
