Compressed Sensing of EEG for Wireless Telemonitoring with Low Energy Consumption and Inexpensive Hardware (1206.3493v3)

Abstract: Telemonitoring of electroencephalogram (EEG) through wireless body-area networks is an evolving direction in personalized medicine. Among various constraints in designing such a system, three important constraints are energy consumption, data compression, and device cost. Conventional data compression methodologies, although effective in data compression, consumes significant energy and cannot reduce device cost. Compressed sensing (CS), as an emerging data compression methodology, is promising in catering to these constraints. However, EEG is non-sparse in the time domain and also non-sparse in transformed domains (such as the wavelet domain). Therefore, it is extremely difficult for current CS algorithms to recover EEG with the quality that satisfies the requirements of clinical diagnosis and engineering applications. Recently, Block Sparse Bayesian Learning (BSBL) was proposed as a new method to the CS problem. This study introduces the technique to the telemonitoring of EEG. Experimental results show that its recovery quality is better than state-of-the-art CS algorithms, and sufficient for practical use. These results suggest that BSBL is very promising for telemonitoring of EEG and other non-sparse physiological signals.

• The paper introduces a BSBL-BO approach that significantly improves EEG signal recovery quality compared to conventional CS methods.
• It leverages compressed sensing to lower energy consumption and hardware costs in WBAN-based wireless telemonitoring.
• Experimental results using NMSE and SSIM metrics, with SSIM values up to 0.85, validate the method’s effectiveness in preserving signal fidelity.

Compressed Sensing of EEG for Wireless Telemonitoring with Low Energy Consumption and Inexpensive Hardware

This paper addresses the technical challenges in wireless telemonitoring of electroencephalogram (EEG) signals through Wireless Body-Area Networks (WBANs). It primarily focuses on three critical constraints: energy consumption, data compression, and hardware costs. The authors propose utilizing Compressed Sensing (CS) in conjunction with Block Sparse Bayesian Learning (BSBL) to effectively manage these constraints, particularly in the non-sparse signal domain of EEG data.

The authors argue that conventional data compression techniques, such as wavelet-based methods, fail to adequately address all three constraints simultaneously. Specifically, these traditional methods often consume significant energy and fail to reduce the device cost to an extent viable for personalized telemonitoring. Compressed Sensing, however, emerges as a promising alternative due to its ability to achieve significant energy savings and maintain competitive compression ratios.

The paper highlights the difficulty of applying CS to EEG signals directly due to their non-sparsity in both time and transformed domains. Unlike traditional CS algorithms that falter under such conditions, the proposed BSBL framework leverages the block structure in data to overcome these challenges. BSBL uses a bound-optimization-based approach which improves the recovery quality of EEG signals beyond that of existing CS algorithms.

Experimental evidence supports the assertions of the paper. EEG data, often replete with signals that lack sparsity in conventional domains, were shown to be faithfully reconstructed using BSBL. Quantitative evaluation using Normalized Mean Square Error (NMSE) and Structural SIMilarity index (SSIM) demonstrates the efficacy of BSBL-BO (Bound-Optimization). The BSBL-BO achieved significant improvement in recovery quality when compared to other CS methodologies, with average SSIM values reaching up to 0.85 for DCT-based BSBL-BO and a notably lower NMSE.

The paper has important implications for practical applications in cognitive neuroscience and other areas where continuous and unobtrusive monitoring of EEG is required. By demonstrating that EEG signals can be effectively compressed and transmitted with low energy consumption using inexpensive hardware, the authors open pathways toward more feasible and economically viable telemonitoring solutions. The research insights into the use of BSBL for EEG compression could inform the design of efficient algorithms capable of handling non-sparse physiological signals, an area traditionally problematic for CS techniques.

Going forward, further investigation into optimizing dictionary matrices or exploring alternative model-based compressive sensing techniques may enhance recovery quality without exacerbating energy consumption. The exploration of hybrid methods combining CS with advanced machine learning techniques could also provide new avenues for improving telemonitoring systems within the constraints of WBANs.

In conclusion, the paper makes a compelling case for the use of BSBL in overcoming inherent challenges posed by EEG signal characteristics, marking a significant contribution to the field of biomedical signal processing for telemonitoring systems. Future research can draw from this work to enhance methodologies and implementations, thereby broadening the scope and applicability of wireless telemonitoring technologies in healthcare.