Research On The Applications Of Compressed Sensing In Communications

Compressed sensing(CS) theory indicates that signal can be sampled under the Nyquistsampling rate. The sampling rate can not be restricted by maximum frequency component of thesignal, but only related to the signal structure. Under the condition that the signal is sparse in thetransform domain or compressible, signal sampling does not need to follow the traditional samplingmode, which firstly sample and then compress. Therefore small amount of data is sampled throughthis way, the sampling devices and storage devices can be liberated.With the development of communication technology, broadband communications and highfrequency applications bring a huge challenge to sampling equipment, so this paper mainly studiesthe application of CS in communication.In this paper, a recognition algorithm for modulation types based on sparse representation isstudied first. Because the traditional recognition algorithm for modulation types based on wavelettransform needs symbol synchronization and large amount of data, result in heavy computationburden, it is difficult to satisfy the requirement of real-time under the battlefield. To solve theseproblems, wavelet transform is used to extracte the transient of different modulated signals.Because the coefficients of wavelet transform have different characteristics whether the signal isnormalized or not, the sparse features of the signal based on wavelet transform are used as theclassification characteristic parameter, and the classification decision process is optimized. Thisalgorithm does not need symbol synchronization, has low complexity, and its recognitionperformance is better than that of the traditional method based on wavelet transform.Secondly, the wideband spectrum sensing based on compressed sensing is researched. Thecurrent researches of spectrum sensing based on CS almost assumes that the sparsity is known, infact, it is unknown and time-varying. Aiming at these problems, a sparsity adaptive algorithm forwideband spectrum sensing is proposed. The distributed compressed sensing and restricted isometryproperty principle are adopted to estimate an initial sparsity value, then the confidence coefficient isused to update the sparsity and the spectrum support set is obtained. Simulation results show thatthe proposed method has better spectrum detection performance than the spectrum sensing methodwith a known sparsity in low SNR.At last, the wideband spectrum sensing based on1-bit compressed sensing are studied. In thedistributed spectrum sensing network, the transmission of sensing information occupiescommunication bandwidth to some degree. A spectrum algorithm based on1-bit compressedsensing and distributed model is proposed. The sensing data of the nodes is quantified in1-bit, i.e. retaining only the symbol of the data. The model of data transmition and storage is simplified, andthe pressure of communication bandwidth is reduced. The simulation results show that thealgorithm is effective.