Compressed Sensing (CS) is a novel technique for data acquisition and Signal detection and is recently applied in various technical fields. Under certain conditions CS allows to solve underdetermined systems of equations correctly. CS is based on the minimization of the l-1 norm which differs from the classical l-2 norm minimization.
The main focus of this research is the application of CS techniques in communication and information technology. In principle, CS allows the reconstruction of signals which are sampled with sub Nyquist frequency. The main prerequisite is that the signal is sparse and contains a certain number of zeros. In a communications framework, CS can be utilized for the detection of signals in sensor networks where nodes transmit data to a central aggregation node. With a low activity probability, the resulting multi-user constitutes a sparse signal. The task of the detector is thus to determine the data and the activity of the nodes in the system jointly. This focus augments current communication systems by a joint activity and data detection leading to a reduced signaling overhead for the system. The main core of this project is the investigation of novel algorithms for the detection of sparse signals and the implementation into a hardware demonstration setup. Besides testing and verification of the algorithms, the tradeoff between performance and complexity is investigated as well.
Starting with a sphere detector for the Maximum a posteriori detection of sparse signals, other sub optimal but less complex algorithms are investigated. The theoretic performance is evaluated with a Code Division Multiple Access (CDMA) network. However, the results of this project can be applied to a various other communication scenarios and form the basis for future CS based detection methods.
The figure shows the solution of an underdetermined systems of equations as the intersection of a line and an Octahedron;