Compressive Sensing Multi-User Detection with Block-Wise Orthogonal Least Squares

One challenging future application in digital communications is the wireless uplink transmission in sensor networks. This application is characterized by sporadic transmissions by a large number of sensors over a random multiple access channel. To reduce control signaling overhead, we propose that sensors do not transmit their activity states; instead sensor activity is detected at the receiver. As sensors have low activity probabilities, the multi-user vector is in general sparse. This enables Compressive Sensing (CS) detectors to perform joint Multi-User Detection (MUD) of activity and data, by exploiting the sparsity. Since sensors are either active or inactive for several symbol durations, block-wise CS detection can be applied to improve the activity detection. In this paper, we introduce blockwise greedy CS MUD, compare it to symbol-wise greedy CS MUD, and show that statistically independent channels for each symbol further improve the activity detection for block-wise CS detection. Herein, we use Code Division Multiple Access (CDMA) as a multiple access scheme.