基于DCT自适应量化的频谱数据有损压缩算法

Lossy Compression Algorithm for Spectrum Data Based on DCT with Adaptive Quantization

针对由于我国频谱监测设备数量激增产生了海量频谱监测数据的存储和传输难题,提出一种基于DCT自适应量化的频谱数据有损压缩算法。首先将多帧频谱监测数据组合并经过灰度空间映射形成一幅时频图,后将时频图数据矩阵按列三等分,将等分后的数据作为RGB三通道合成一幅彩色图像并对其进行二维DCT变换,利用BP神经网络估计量化阈值后进行量化,最后经过游程编码与二次熵编码完成频谱监测数据的压缩。通过对真实接收机采集的频谱数据的压缩处理,压缩率约为10%,数据恢复后的百分比均方根误差约为11%。实验结果表明,所提算法具有稳定的压缩率和较小的恢复误差,能够有效地对实际采集的频谱数据进行压缩。

In response to the storage and transmission challenges posed by the increasing number of spectrum monitoring devices inChina, this study proposes a lossy compression algorithm for spectrum data based on the Discrete Cosine Transform (DCT) withadaptive quantization. Firstly, multiple frames of spectrum monitoring data are combined and mapped to a time-frequency imageusing grayscale mapping. Then, the data matrix of the time-frequency image is divided into three equal parts column-wise, whichare treated as RGB channels to compose a color image. The image is then subjected to two-dimensional DCT transformation.Subsequently, the quantization threshold is estimated using a BP neural network and applied for quantization. Finally, thecompressed spectrum monitoring data is achieved through run-length encoding and second-order entropy coding. Experimentalresults on real-world collected spectrum data show a compression ratio of approximately 10% and a percentage root mean squareerror of about 11% for the recovered data. The results demonstrate that the proposed algorithm exhibits stable compressionperformance and low recovery error, effectively compressing the acquired spectrum data in practical scenarios.