| Wireless sensor systems are limited by the energy consumption of the sensor nodes, which are themselves limited by the transmitter. Transmitter energy consumption can be reduced by reducing the amount of transmitted data, which is usually determined by the maximum bandwidth of the signal of interest according to Nyquist theory. Compressed sensing is a mathematical framework for sampling signals at the rate of information rather than at the Nyquist rate. If the signals of interest are su ciently sparse in some domain, compressed sensing can be used to reduce the number of samples required to reconstruct the signal on the receiver side to below that dictated by Nyquist theory. Thus, compressed sensing can reduce power consumption by compressing data at the source. This thesis presents the analysis and experimental results of a compressed sensing based analog-to-information converter (AIC) in 90nm CMOS technology. This AIC utilizes a random sampling analog-to-digital converter (ADC) to acquire the data, and l1-minimization to reconstruct the data. It was found that this algorithm performs poorly as sparsity increases, and a modied algorithm that exploits group sparsity was used to demonstrate acquisition of wide band signals as well as various biomedical signals. Applicability of these methods are analyzed in depth in particular for wireless acquisition of multi-channel EEG. |
