MP3编码量化模块的改进

MP3是应用最广泛的一种音频格式，MP3解码芯片已发展成熟，却很少有实时编码器产品。这是因为MP3编码算法的运算量远大于解码算法，而量化模块在编码算法中占很大的比重(38．2％)，为实现便携式设备的MP3实时编码，有必要对量化模块进行改进来降低运算复杂度。在分析了现有的一些改进方法后，提出了在外循环中限制量化频带以滤去高频信号，计算比例因子初值以减少循环次数；内层循环中计算量化因子初值后自适应调整步长的方法。在获得良好音频质量的同时将量化模块运算量减少到标准算法的6％，使得整个编码算法的运算量降低到标准算法的64％，配合简化声学心理模型和分析滤波器组快速算法，可在单片DSP或ARM上实现MP3实时编码。

An Enhanced MP3 Encoder Quantization Algorithm

MP3 is the most popular audio format nowadays, although MP3 decoder has been widely used, there are few successful real-time encoder available. This results from that MP3 encoder is much more complicated than the decoder, and the quantization module consumes a large amount of computation. In order to realize a real-time MP3 encoder that can be used in portable devices,the complexity of quantization module should be greatly reduced. After analyzing some available improvements, an enhanced algorithm is proposed, which can limite scalefactor band to filter high frequency signals, compute the initial scalefactors to decrease outer iterations and adaptively change global gain after computing its initial value that is close to the ideal output so that inner iterations are greatly reduced. Tests show that this algorithm can not only produce high quality audio outputs, but also minimize the computation load to 6% of the standard realization. Therefore, the whole encoder consumes only 64% resources of the standard encoder. Combining with simplified psychoacoustic model and fast analysis filter bank, a real-time MP3 encoder can be implemented on a single DSP or ARM.