In signal coherent detection, data-aided (DA) method has high detection accuracy, but unavoidably suffers from a large processing delay. Meanwhile, a decision-directed (DD) method incurs no processing delay, but suffers from error propagation problem. Motivated by the above problems of DA and DD methods, the authors present a novel signal detection algorithm for delay-sensitive applications employing orthogonal frequency-division multiplexing systems. The proposed detection method provides similar detection accuracy to the DA method and incurs no processing delay. This method employs channel estimation followed by linear prediction and interpolation to obtain the channel-state information for symbol detection. To achieve a better prediction performance, they perform the linear prediction on the channel taps in the time domain, and conduct the interpolation using a raised cosine filter. Simulations on physical channel models verify the improvements of the proposed method.