In this paper, we investigate the estimation of fast-fading LTE downlink channel in high-speed applications of LTE Advanced. A robust channel estimation and interpolation algorithm is essential in order to adequately track the fast time-varying channel response. Using Support Vector Machine Regression (SVR) and exploiting kernel features, we develop an extended algorithm to jointly estimate the complex-valued channel frequency response in time and frequency domains from the transmission of known pilot symbols. This paper integrates both channel estimation at pilot symbols and interpolation at data symbols into the complex SVR interpolation method under very high mobility condition. The Bit Error Rate (BER) and Mean Square Error (MSE) performances of our fast-fading channel estimation scheme is demonstrated via simulation for LTE Downlink with 64-QAM modulation and 500 Km/h velocity.