Quantization effects in an analog-to-information front end in eeg telemonitoring

Energy in wireless communication is the dominant sector of the energy consumption in electroencephalography (EEG) telemonitoring due to intrinsically high throughput. Analog-to-information conversion, i.e., compressed sensing (CS), offers a promising solution to attack this problem. Most of previous research work on CS focus on the sparse representation to reduce the signal dimension, but the impact of quantization in CS has had limited examination in the research community. In this brief, we investigate the quantization effects of CS with the application in EEG telemonitoring. In particular, we study the quantized CS (QCS) structure to explore the impacts of quantization on the performance-energy (P-E) tradeoff of the front end in EEG telemonitoring. Compared to the state-of-the-art CS with the constant bit resolution, experiments show that the QCS framework with the optimal bit resolution can improve the P-E tradeoff by more than 35%. Furthermore, the optimal bit strategy even broadens the application range of the QCS framework by 54% compared to the traditional Nyquist sampling, which indicates that the quantization is a critical factor in the entire CS framework.