| Accurate estimation of the Doppler spread, or the mobile speed, which indicates the rate of wireless mobile channel variations, is important for many applications such as handoff, adaptive modulation and equalization, power control, etc. There are three major classes of speed estimation techniques: crossing-based methods, covariance-based methods, and maximum likelihood (ML) based methods. Crossing-based approaches rely on counting the number of received signal's level crossing which is proportional to the mobile speed, while covariance-based algorithms exploit the maximum Doppler frequency information which exists in the sample autocovariance of the received signal. However, both crossing-based and covariance-based speed estimators are sensitive to noise, especially for small Doppler spreads. Although the ML-based estimators are optimal or near optimal, they are complex to implement, need the knowledge of signal-to-noise-ratio (SNR), and require noise to be Gaussian. Compared with these conventional speed estimators, the technique we proposed, which is based on the power spectrum density (PSD) of fading channel, is not only robust against the noise, including both the Gaussian and impulsive non-Gaussian, but also insensitive to nonisotropic scattering observed at the mobile station (MS). In this paper, we extend the temporal-only speed estimator of the PSD-based method to a spatio-temporal estimator, using multiple antennas at the base station (BS). By taking advantage of the spatial information provided by an antenna array, the new estimator performs very well over a wide range of noise power, nonisotropic scattering, and line-of-sight component, verified by extensive Monte Carlo simulations. |