In general, frequency domain based techniques are preferred to time domain techniques for the implementation of audio crosstalk cancellation due to time domain techniques suffer from more computations. In this paper, the computational difficulty of real FFT on DSP processors is analyzed and to overcome the disadvantages, two optimization methods (1. Parallel Filtering, 2. Mixed Filtering) are proposed for the implementation of Audio Crosstalk cancellation. These methods are developed based on fast convolution using overlap save method & Finding DFTs of two real sequences using single Complex DFT. By applying these properties to two real outputs of stereo crosstalk cancellation structure, valid equations are derived for implementation. The proposed methods are simulated in real-time on 32 bit floating point processor. The results show that the proposed methods can significantly reduce the computational complexity when compared to conventional time domain convolution & frequency domain based individual filtering without affecting the performance.