| IntroductionOrthogonal frequency division multiplexing (OFDM) is an attractive transmission method for high data transmission because of its excellent bandwidth efficiency and good immunity to multi-path fading and impulse noise. Owing to its desirable advantages, it has been adopted as standard for digital audio broadcasting, digital video broadcasting and indoor wireless LAN.If we want get the benefit form OFDM systems, a series technology need to be carried out. Channel estimation plays an important role in OFDM systems, which is essential to bit and power allocation and signal detection. Without perfect knowledge of channel state information, the OFDM systems either cannot work or may incur significant performance loss.OFDM channel estimation methods can be divided into three kinds. The first one is named Non-Blind Channel Estimation (NBCE), a channel estimation method based on pilot or trained sequence. The advantage of this algorithm is easy to realize, but it occupy many spectrum of the systems; The second is blind channel estimation (BCE). It based on the limited characters of the transmitted information symbols and their statistical trait. Although it does not need pilot symbols or trained sequence, saving the bandwidth, but it converges slowly and has high complexity. The third class is the semi-blind channel estimation (SBCE) by using the information from blind channel estimation algorithm and known sampling symbols to finish channel estimation. It solves the problems of spectrum waste from channel estimation based on pilot symbols or trained sequence and high complication of blind channel estimation methods. So the semi-blind channel estimation algorithm is regarded to be a promising way for channel estimation.A new semi-blind channel estimationIn this paper, we present a semi-blind channel estimation method for OFDM using PSK and QAM signaling. This is a new semi-blind channel estimation method for OFDM systems. Since it's a semi-blind estimation, it has both the advantages of BCE and NBCE and the algorithm can be divided into two steps. First, we arrange few pilots which is nonuniformly spaced to get part channel impulse response. Linear minimum mean-squared error (LMMSE) estimates of the channel gains over the pilot subcarriers are obtain by backrotating the received signal according to the knowledge of the pilot symbols. Then, the LMMSE estimates are interpolated over the entire frequency grid. This is the initializedchannel response estimation vector for the following channel estimating..Though a LMMSE estimator in the first step using only frequency correlation has lower complexity than one using both time and frequency correlation, it still requires a large number of operations. We introduce a low-complexity approximation to a frequency-based LMMSE estimator that uses the theory of optimal rank reduction. By using the singular-value decomposition (SVD), an optimal low-rank estimator is derived, where performance is essentially preserved.In fact, not only the pilots provide the channel information, the data subcarriers also give the information about the magnitude response of the channel. So in the next step we can define a objective function for an OFDM frame, which is a sum squared error for pilot signals and a sum of weighted squared generalized magnitude error for data signals in frequency domain. It is found that the sum of squared absolute magnitude error for data signals provides an excellent block mean square error (BMSE). An efficient iterative algorithm for obtaining an optimal solution is derived, a differential action is performed, it can be learned that only the channel phase vector needs to be updated in the iteration. After several iterations, a channel estimation vector which is close to the real vector is obtained. That takes several iterations to converge to the steady state solution. The new algorithm is shown to provide a better performance gain than other pilot-based channel estimation methods and semi-blind methods in bad channel environment.The advantages of this algorithmCompared with literature [1], main advantages in this paper are summarized as following:1) In practice, there exists a bad channel environment which has lots obstacles, so the signal may arrive form as high as 100 paths, which means the channel has long impulsion response, even longer than the pilots'numbers. In the literature [1], its NBCE parts use MLE method to obtain initialized channel response estimation vector. Under this condition MLE method is invalid, and leads the BCE part failed too. In order to overcome these deficiencies, this paper proposes an improved semi-blind estimation algorithm based on LMMSE and its simplified method. Since the improved method use the channel statistics, it can overcome the deficiencies of literature [1], and can provides good performance even under the bad channel environment.2) Even if the channel environment is not so bad, we can also use the improved method to obtain high system performance. As known to all, the main shortcoming of the NBCE method is its waste system resources. With the semi-blind method based on the LMMSE, we can reduce the pilots'numbers, even reduce to the number fewer than the length of the channel impulsion, so we can use more subcarriers to transmit the data, increase the utility of the bandwidth.3) In the paper, we arrange the pilots with nonuniform space. From literature it is found that if there is no suppressed carriers, the best performance is achieve by resorting to uniformly spaced pilots with separation interval. Unfortunately, such a condition is not always met in practice. When some carriers at the edges of the spectrum are suppressed, the only way to determine the optimal pilots'locations is through exhaustive search. However, some qualitative conclusions can be drawn: the minimum block mean square error can be achieved with decrease the pilots'distance at the edges.4) Literature [1] and this paper based on different assumptions about the channel impulse response. In the former, use MLE to estimate the initialized channel vector, the channel impulse response is viewed as a deterministic but unknown vector. whereas in the latter, it is regarded as a random vector whose particular realization we want to estimate. Correspondingly, the mean squared error in the literature [1] is understood as an average over the observed data, whereas in this paper, the average is taken not only over the data but over the channel impulse response probability density function as well. It follows that the method in this paper has the minimum MSE"on the average", i.e. with respect to all the channel impulse response realizations.Conclusions and Future WorkIn this paper, a new single frame-based semi-blind channel estimation method has been presented. For this method, using LMMSE algorithm in the NBCE part, and then the generalized magnitude error term for data signals incorporated with pilot signals is proved to yield a channel estimate giving better BMSE than algorithm for literature [1] in a bad transmit environment. The new semi-blind method in this paper is shown to give good performance in regard to bit error rata (BER) and also have an efficient implementation. Unlike most of subspace-based semi-blind methods that needs many OFDM blocks for channel estimation, the new method just takes one single OFDM block for carrying out estimation, which enables it to perform in fast fading channel.The channel excess delay is usually assumed to be equal to integral sample, however, this assumption is seldom satisfied in practice. In that case, it will emerge energy leakage. In the future, we have to pay attention to the channel excess delay of non-integral sample time and try to solve this problem in further study. |
PDF Full Text Request