高效的MPEG先进音频编码方案

Efficient Design of MPEG Advanced Audio Encoder

为提高MPEG先进音频编码系统的编码效率,分别在三个关键模块上进行了算法优化,提出一种高效的编码实现方案.在心理声学模型中,使用新的时域分块峰值变换率准则代替感知熵来判断MDCT变换块的类型,降低误判、漏判概率,提高编码质量和速度;在分析滤波器中,基于双路并行计算技术,采用N/8点FFT算法实现N点MDCT变换,提高运算速度;在量化编码模块中,利用量化噪声能量公式,减少量化迭代次数,提高编码效率.该编码方案在保证音频质量的前提下,减少了50%的编码时间,满足实时性系统设计的要求.

An efficient audio encoder for MPEG AAC is exhibited in this paper, by carrying out the algorithm level optimizations of three critical modules. In psychoacoustic model,a block switching method in time domain is presented instead of PE-based algorithm ,in order to resolve the quality degrade caused by erroneous decision of block type. In addition, an analysis filter calculating the N-point MDCT transform employs a parallel scheme basing on N/8-point FFT, improving the computation rate and the data throughput. Finally, to meet the bit rate/distortion critical,a novel technique through pre-processing and adjustment of scale factor in quantization block is used. According to the objective evaluation results,the encoder proposed in this paper gets the performances of both low complexity and high audio quality.