PsychoacousticModelforMPEGAAC.PDFVIP

Memory and Computationally Efficient Psychoacoustic Model for MPEG AAC on 16-bit Fixed-point Processors Shih-Way Huang, Liang-Gee Chen Tsung-Han Tsai Grad. Inst. of Elect. Eng. and Dept. of Electrical Engineering, Dept. of Electrical Engineering National Taiwan University, National Central University, Taipei, Taiwan, ROC Chung-Li, Taiwan, ROC shihway@video.ee.ntu.edu.tw han@ee.ncu.edu.tw Abstract—When MPEG AAC encoders are implemented on 16-bit fixed-point processors, the constraints of low complexity II. PAM REVIEW and limited data word length become challenges especially on Fig. 1 illustrates the block diagrams of MPEG AAC the critical algorithm, Psychoacoustic Model (PAM). This (above) and PAM (below) based on [1]. PAM calculates a paper proposes a very efficient PAM design concerning the masking threshold, which is the maximum distortion energy reduction of memory requirement and computational masked by the signal energy for each coding partition. complexity while maintaining the audio quality. Moreover, it also predicts a block type, which determines the block length used in other parts of the encoder. Meanwhile, I. INTRODUCTION Modified Discrete Cosine Transform (MDCT) transforms input audio samples in time domain into spectrums in Portable electronic devices with audio playback and frequency domain. The frequ