论文部分内容阅读
数据压缩的方法很多,实际应用中多采用变换加编码的方法,在允许一定的误差的范围内可以获得比无损压缩高得多的压缩率,而且常常大大简化处理算法。采用一种分块的动态归一化将需要压缩的数据收缩到[-1,1]的区间内,再采用Llyod算法对归一化的数据进行非线性标量量化编码降低每个采样点的比特位宽。算法简单,易于硬件实现,解码时只需查找码书和动态恢复。在50%压缩比情况下EVM值在1%以内。并针对该算法进行了MATLAB仿真和硬件代码的编写。
There are many methods of data compression. In practice, many methods of transform and coding are used, which can obtain a much higher compression ratio than lossless compression within a certain range of error, and often greatly simplify the processing algorithm. A block of dynamic normalization is used to compress the data to be compressed to the range of [-1,1]. Then, the Llyod algorithm is used to perform non-linear scalar quantization on the normalized data to reduce the bit of each sample Bit wide. The algorithm is simple, easy to implement in hardware, only need to find the code book and dynamic recovery when decoding. The EVM value is within 1% at 50% compression ratio. Aiming at this algorithm, MATLAB simulation and hardware code are written.