Subspace-based Multiuser Detection For DS-UWB

1. Research meaning of thesisThe detection technique of signal is a focal point in the UWB field. Four main schemes have been brought up, which are RAKE receiver, channel estimation, TR receiver and multiuser detection. The problem of RAKE is that it has to make a choice between complexity and performance. There are two ways to estimate the channel. Channel estimation with data-aided estimates fast on the cost of spectral utilization and power utilization. Blind channel estimation which does not need trainning sequence has the disadvantages of high computing complexity and slow convergence speed. TR receiver which employs the received signals having noise as the reference signals for data detection will lead to significant performance degradation. Among these ways, multiuser detection which performs well has high computation complexity and needs to be more practical. At present, some researchers are looking for suitable multiuser detection methods for UWB. Thesis proposes a scheme to detect DS-UWB signals based on subspace, and SIMULINKis used to test the schems's performance and verify the feasibility of the scheme in the end.2. Main content of the thesis(1) Research on UWBUWB transmits information by impulse with period of nenoseconds. The pulse wave determines the distribution of the frequency energy. First, some typical pulse wave are discussed and simulated. Second, some UWB modulation schemes are discussed. There hasn't been a standard for the UWB system yet. MB-OFDM and DS-UWB are compared. DS-UWB is choosed to be the system scheme, inwhich the multiuse detection is applied. The system which doesn't adopt all the setting is simplified for simulation convenience.UWB channel model is not the same as what is used before in the traditional communication technique. Some typical channel models are choosed to study. IEEE802.15.3a is adopted to be the simulation channel.(2) Choice of subspace trackingSubspace method is a kind of blind multiuser detection. It projects the received signals into the signal space and noise space, and also projects the filtering coefficient of MMSE into the signal space. The subspace estimation and subspace tracking are simplified by subspace algorithm. The MMSE blind adaptive multiuser detection can be applied with adaptive method based on signal subspace tracking whose performance influences the detecton directly. Two sorts of algorithms, which are one-rank perturbation and NAHJ-FST, are chosen to discuss and simulate, because the point of thesis is not to study how to improve the subspace tracking. Finally, NAHJ-FST algorithm is chosen for its low complexity and no needs to choose the exact inital value.(3) Simulation and analysisAfter proposing the scheme and studying the relevant part, thesis simulates the whole system with software SIMULINK. The simulation is divided into transmitter, channel and receiver.The transimmiter is set as follows: First, user's datas are sent into the differential encoder, and then the BPSK modulation with phase pi/2 is done. Second, the signals are spreaded with the KASAMI code in length of 15.The wave is the 2nd derivation of Gauss pulse in width of 0.3ns and period of 1ns. Finally, the signals are sent into the UWB indoor multipath channel after having been added the white Gauss noise.The channel model proposed by INTEL is adopted. The simulation mainly runs in the CM1 and CM3 channels.The UWB indoor multipath channel model is a time-varing model, which is complex in the simulation, and therefore is set to be fixed. All the signals pass the same channel.The receiver is setted as follows: Signals are first sent into the match filter to get the vector at instant k, and then the vectors are sent into the subspace tracking block to get the eigenvector subspace and the eigenvalue subspace with fast subspace tracking algorthm. Meanwhile, the channel is estimated with eigenvector subspace and eigenvalue subspace. Finally, the MMSE detector vector is obtained, which will be decided and sent into the decoder to decode. The datas out of the differential decoder are the final ones we want.The simulations are run to test the influence on the system performance mainly in five situations which are the length of the data, near-far effectiton, user numbers, channel and encoding. The simulation results show that the scheme proposed works well in the CM1 channel but badly in CM3 channel. There are three reasons as follows:a) Subspace tracking algorithm. At first, the subspace algorithm is proposed aiming at the sparse multipath. This algorithm with good performing has low complexity on the cost of not strong orthogonality. Meanwhile, it cann't correct the deviation in time for being a blind adaptive multiuser detection results in its poor performance in the more complex multipath channel.b) Channel estimation algorithm. The channel estimation algorithm is mainly used to estimate the equivalent spreading codes of the expected user. The KALMAN algorithm is chosen. Choosing an algorithm with better performance will lead to getting more accurate equivalent spreading codes, which will improve the system's performance in the strong multipath channel.c) Spreading codes. KASAMI codes with length of 15, which are not strong orthogonal and whose length is short, are choosed. Choosing a long spreading code with better orthogonality, which won't create the complexity of the system, will improve the system's performance.(4) Improving schemesTwo improving schemes are proposed as follows:a) Adopting the subspace algorithm with structure improved. A RAKE structure, which collects the releavant user's signal in each path, is added to thefront of the subspace detector. The branch detector is solved with the subspace analysis. Then all the collected signals are decided combined in the integrated way to depress the multipath influence to get the expected user's signal.b) Adopting the group blind subspace algorithm. The subspace algorithm only knows the expected user's spreading codes. The group blind subspace algorithm does not only know that but also the other user's knowledge.3. Research findingsThe scheme's performance is tested with a series of simulation, the result of which shows that the system works well in the CM1 channel but badly in the CM3 channel.Considering that the system is built on a rough situation and not added the relevant protection and correction, the schems proposed is feasible.