Design and Implementation Analysis of OSD based Audio Crosstalk Cancellation with Multi-Channel Inputs on DSP Processors

Abstract

To overcome the basic system inversion problems in conventional crosstalk cancellation systems, the three channel Optimum Source Distribution (OSD) was developed by Takashi et al., that provides balance between in-phase and out of phase components. The main difficulty in this system is real-time implementation of cross talk cancellation filters on DSP platforms, particularly when filter lengths are very long with system operates on multichannel inputs. This paper discusses the implementation complexities and proposes an efficient approach to reduce the power consumption. It also discusses efficient usage of available processor memory. By utilizing the processor resources and existing frequency domain techniques, this approach would be one of best methods for long filters. The mathematical model, with efficient design for floating point DSP processors, clearly explains the optimization methods at algorithmic and instruction levels. The computational complexity of the proposed method was measured for various multi-channel input sources and the comparison was shown. The results indicate that the proposed method provides efficient computations than existing methods. Keywords: Convolution, Crosstalk Cancellation, Optimal Source Distribution, Overlap Save Method, Uniform Partitioned Convolution