Channel Estimation for Communications
This work has been supported by NSF and ONR
In our multimedia era we are witnessing an increased demand for high speed digital transmission; a direct result of the need for fully integrated digital voice, image and data services, and high speed radio indoor and mobile communications. One of the major obstacles to high signaling rate digital communications is sensitivity to Intersymbol Interference (ISI), caused by non-ideal dispersive channels and multipath propagation. ISI degrades the performance and imposes limitations on transmission rate. The development of compensating schemes for ISI (namely, adaptive equalizers) is of ultimate importance in achieving high transmission rates. We are developing multichannel equalization schemes that require no or minimum statistical knowledge about the transmitted signal.
Cross-Spectrum Based Blind Channel Identification and Equalization
It is well-known that the analog output of a linearly modulated communications signal is cyclostationary, and, as such, its second-order statistics contain phase information. This fact has been used implicitly in the recent years in order to develop algorithms for the blind identification of the unknown propagation channel dynamics. We have developed a technique that explicitly uses phase information in order to identify the unknown channel.
The proposed method assumes a single-input two-output scenario, which can be obtained either by sampling the output of two antennas (receivers) at the transmitted symbol rate, or by over-sampling a single antenna output at a twice higher rate. Then the phase of the cross-spectrum of these two observations, after they have been filtered through their minimum-phase equivalent sequences, is estimated. Based on discrete samples of this phase sequence, and using the theory of system reconstruction from phase, an over-determined linear system of equations can be formed. The solution of this system is a sequence which contains combinations of the minimum and maximum-phase parts of the unknown channels. This sequence can be decomposed into the minimum and maximum-phase parts of the channels, from which the actual channels can be obtained.
A combination of the above routines, and the implementation of a desired number of Monte-Carlo simulations, is achieved by using the function automc.m. In order to generate ARMA variates corrupted by additive Gaussian noise, iterate.m uses the functions PAM.m and real_noise.m (M-PAM case), or qam.m and addnoise.m (QAM case). If the input is an M-PAM, or QAM process, then a calculation of the Mean-Square-Error (MSE) and Symbol-Error-Rate (SER) is performed, using the function calc_fse_SER.m, which in turn calls MMSE_fse.m and arma_fse.m. Examples of representative channels for communications applications are generated using the functions ray2.m, ray3.m and rcosine.m.
For more details on the system identification methodology and several issues pertaining to its implementation, the reader is referred to  and .
Key papers (Channel Estimation for Communications)
R. Lin and A.P. Petropulu, “Linear precoding assisted blind channel estimation for OFDM systems,” IEEE Trans. on Vehicular Technology, vol. 54, issue 3, pp. 983-995, May 2005.
A.P. Petropulu, R. Zhang and R. Lin, “Blind OFDM Channel Estimation through Simple Linear Precoding,” IEEE Transactions on Wireless Communications, vol. 3(2), pp. 647-655, March 2004.
S. Yatawatta, A.P. Petropulu and R. Dattani, “ Blind Channel Estimation Using Fractional Sampling ,” IEEE Trans. on Vehicular Technology, vol. 53(2), pp. 363-371, March 2004.
Key papers (Cross-Spectrum Based Blind Channel Identification and Equalization)
Haralampos Pozidis and
Athina P. Petropulu, "Cross-Spectrum Based Multichannel
Blind Identification," IEEE Transactions on Signal Processing, vol.
45, no. 12, pp. 2977-2993, Dec. 1997.
Haralampos Pozidis, "Techniques for Blind Channel Identification and Application to the Equalization of PAM/QAM Modulated Signals", Ph.D. Dissertation,