Study On Channel Estimation Algorithm In MB-OFDM UWB Communication System

In recent years, ultra-wideband (UWB) technology has gradually become a hotspot in the research and development of wireless communication area, and has received a significant amount of attention from academic and industry, as well as one key technology of next generation wireless communication technologies. UWB technology becomes more and more attractive because of its advantages, such as high transmission rate, low power consumption, low cost, robustness to multi-path fading capability and high fine timing precision. But traditional communication systems can not compare with this system. UWB communication can be divided into two realization methods, one is impulse radio, and the other is carrier modulation. Now there are two technical solutions: direct sequence code division (DS-CDMA) and multi-band orthogonal frequency division multiplexing (MB-OFDM). MB-OFDM UWB is superior to DS-CDMA UWB in bandwidth allocation, anti-jamming and transmission rate.Channel estimation is the basis of correlation detection, demodulation and equalization. So study on channel estimation has important significance. There are three kinds of channel estimation technology. The first one is non-blind channel estimation, the second is semi-blind channel estimation and the third is blind channel estimation. In this thesis we focused on channel estimation in MB-OFDM UWB system based on training sequence (non-blind channel estimation). Least square (LS) estimation algorithm and DFT-based channel estimation algorithm have been analyzed. Through making use of physical layer structure of MB-OFDM UWB communication system, based on the average of the LS channel estimation algorithm is proposed and a way to make use of time redundancy property to modify LS channel estimation algorithm is proposed. Because DFT-based channel estimation algorithm within the rectangular window still contains the noise, in order to reduce noise based on most significant taps (MST) channel estimation algorithm is proposed. This method can not only reduce noise but also improve the performance of the system, and the computing complexity is reduced. And the channel estimation algorithm in the time domain noise can be reduced using Wiener filter.At first, we introduced the LS estimation algorithm and its improved algorithms. In the same bandwith sending two same training sequences for channel estimation, based on the average of the LS estimation, an algorithm is proposed and it can improve the performance of system. Subsequently, making use of time redundancy (TR) to improve the LS estimation algorithm is proposed. The algorithm sends data only to the odd or even number sub-carriers and the corresponding even or odd sub-carriers are set to zero. In the receiver, linear interpolation algorithm in frequency-domain is applied due to its simplicity, applicability and efficiency for a relatively flat fading channel to obtain the channel information of the sub-carriers without data information. Then the whole channel frequency response can be obtained. This proposed channel estimation algorithm is called TR with linear interpolation (TRLI). In the channel CM1, the performance of this algorithm is superior to traditional LS estimation algorithm. But in channel CM4, a strong frequency selective fading channel, using TREO, channel estimation algorithm is proposed. This algorithm can better resist frequency selective fading. Because channel gain between two adjacent carriers has small correlation in the strong frequency selective fading channel. This algorithm is better than TRLI channel estimation algorithm. The simulation results show that the performance of the improved LS estimation algorithm is superior to the traditional LS estimation algorithm.LS channel estimation is simple, but it has poor performance. In LS estimation there is no use of the correlation between frequency domain and time domain. DFT-Based channel estimation algorithm can improve system performance. In MB-OFDM UWB system, FFT module has already used, and computing complexity is not increased. This algorithm only in the time domain adds a simple rectangular window in order to reduce the noise outside of the window, and within the window still contains the noise. Although it has minimal complexity and is easy realized, the performance is not satisfied. In order to reduce the noise, the algorithm in time domain using Wiener filter to reduce noise is proposed. And a time domain Wiener filter coefficient matrix is derived. It can maximally reduce the impact of noise with very large complexity and computation, and it is difficult in practical use. A MST-based algorithm is proposed to further reduce the effect of noise within window. The algorithm makes a compromise between computing complexity and performance. It further processed the noise within rectangular window according to the mean square error criterion and a best threshold is derived. If the channel energy is less than this threshold value, it will be set to zero. But ignoring low energy channel components can degrade system performance and also can eliminate the noise impact from those components. Generally speaking, noise from those channel components neglected are higher than the contained energy they, especially in low SNR. The simulation results show that the performance of MST-based algorithm is better than that of the DFT-based algorithm, and computational complexity is also lower.