基于PLI算法的AAC反量化和PNS解码模块设计  

Design of AAC Inverse Quantization and PNS Decoder Based on PLI method

作　　者：张红升[1,2] 王国裕[1,2] 陆明莹[1] 蒋涛[1] 郭光宇[1] 

机构地区：[1]重庆邮电大学微电子工程重点实验室,重庆400065 [2]西安交通大学微电子学系,陕西西安710049

出　　处：《微电子学与计算机》2012年第8期1-5,共5页Microelectronics & Computer

基　　金：科技部(2009GJF10046);重庆市教委(KJZH10205);重庆市经信委(2009[57]);重庆市人社局(2010[261])

摘　　要：针对AAC(Advanced Audio Coding)解码器中具有复杂运算因而难以硬件实现的反量化模块和知觉噪音替代(Perceptual Noise Substitution,PNS)模块,提出用分段线性插值(Partition and Linear Interpolation,PLI)进行近似计算的方法.其基本原理是将复杂的曲线分割成多个区间,在每个区间内用线性插值的方法进行近似.利用该方法设计的反量化模块和PNS解码模块获得了较高的运算精度,最大运算误差率分别为2e-4和5e-3,全部运算可分解为简单的查表、乘法、加法和移位运算,易于硬件实现.FPGA验证结果表明,反量化模块仅耗用了45个逻辑单元和2 304ROM bits,PNS解码模块仅耗用了21个逻辑单元和144ROM bits,且解码音质良好,没有产生可感知的噪音.A partition and linear interpolation （PLI） method is proposed, in order to simplify the complicated algorithm of the inverse quantization and perceptual noise substitution （PNS） blocks of the advanced audio coding （AAC） decoder. The principle of the PLI method is to divide a complicated function into several regions and apply linear interpolation in each region to obtain the approximate value of the function. The simplified algorithms achieved low max error rates as 2e-4 and 5e-3 for inverse quantization and PNS decoder respectively. The computations comprise only look-up table （LUT）, multiplication, add, and shift operations, thus can be easily implemented in hardware. The FPGA verification results showed the inverse quantization block consumes only 45 logic elements and 2 304 ROM bits, and the PNS decoder consumes only 21 logic elements and 144 ROM bits. The audio quality is good and no perceptible noise is generated.

关 键 词：AAC 分段线性插值 反量化 知觉噪音替代 

分 类 号：TN492[电子电信—微电子学与固体电子学]