SEMG signal compression based on two-dimensional techniques.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Data Compression

Background:

• Surface electromyographic (SEMG) signals are increasingly compressed using 2D techniques and image/video encoders.
• Existing methods often require specialized compressors or preprocessing for optimal SEMG data organization.
• Adaptive frameworks offer a direct approach to tackle SEMG signals represented as images.

Purpose of the Study:

• To propose and evaluate a novel 2D approach for SEMG signal compression.
• To introduce a new preprocessing technique, segmentation by similarity (SbS), to improve correlation exploitation.
• To assess the performance of modified algorithms and off-the-shelf encoders for SEMG compression.

Main Methods:

• A modified multidimensional multiscale parser (MMP) algorithm adapted for SEMG signal redundancies.
• Introduction of segmentation by similarity (SbS) preprocessing for enhanced intra- and inter-segment correlation.
• Evaluation using percentage difference sorting (PDS) with HEVC, H.264/AVC, and JPEG2000 image compressors.

Main Results:

• Dynamic SEMG signals achieved good compression factors with H.264/AVC and HEVC combined with preprocessing.
• The modified 2D MMP algorithm outperformed existing methods for isometric signals at low compression factors.
• SbS with HEVC was competitive at high compression factors; JPEG2000 with PDS offered good performance and low complexity.

Conclusions:

• The proposed SEMG compression schemes are effective.
• The modified MMP algorithm is a viable alternative for isometric SEMG signal compression.
• Off-the-shelf image encoder approaches facilitate rapid implementation and widespread adoption.