Polyphase Parallel Source Layer Architecture for Low Rate Encoding

Abstract

A <inline-formula> <tex-math notation="LaTeX">$L$ </tex-math></inline-formula> -interpolation filter can increase the sampling rate from <inline-formula> <tex-math notation="LaTeX">$f_{s}$ </tex-math></inline-formula> to <inline-formula> <tex-math notation="LaTeX">$L{\,}f_{s}$ </tex-math></inline-formula> that must be managed by increasing bandwidth and data throughput. Consequently, we propose a polyphase interpolator architecture design that can implement the encoder and decoder pair at an effective rate of <inline-formula> <tex-math notation="LaTeX">$2f_{s}$ </tex-math></inline-formula> instead of <inline-formula> <tex-math notation="LaTeX">$L{\,}f_{s}$ </tex-math></inline-formula> by operating directly on a subgroup of the polyphase branches. The computational complexity of the proposed polyphase structure remains reduced by half, since it retains the coefficient symmetry property of the linear phase FIR filter. Rate-distortion analysis demonstrates improved parallel entropy coding without compromising signal fidelity. Moreover, the proposed method achieves progressively lower overhead as the interpolation factor increases, yielding 7.24% and 15.60% average overhead reduction at <inline-formula> <tex-math notation="LaTeX">$L=4$ </tex-math></inline-formula> and <inline-formula> <tex-math notation="LaTeX">$L=8$ </tex-math></inline-formula>, respectively, relative to the representative parallel bitstream model.