A time domain based efficient block decision algorithm for audio coders

In typical audio encoders the block decision is done either using time-domain techniques like energy computation or frequency domain techniques like temporal noise shaping (TNS) [1], [2]. The time-domain energy computation based decisions are less effective for detecting many of the stringent scenarios presented by test cases like castanets and fatboy. The frequency domain based algorithms have better decision making capabilities, however they are inherently complex as they require the computation of the FFT, additionally in case of TNS the computation of LPC (Linear Prediction coding) in the frequency domain. An improved time-domain technique with better block decision capability compared to TNS and with lesser computational complexity is proposed in this paper.