Research On Spectrum Data Compression Method Based On Pattern Recognition

In recent years,with the rapid development of radio monitoring automation technology,the continuous expansion of the construction scale of spectral monitoring network,the long-term and large-scale spectrum monitoring produce mass monitoring data.These monitoring data are not only the valuable resource to study radio spectrum,but also the data support for performance analysis,failure situation recurrence and so on.However,how to quickly process and store the real-time data produced during spectrum monitoring,which is the key to promote the development of radio spectrum monitoring technology.Due to the most of the radio spectrum frequency data are steady state,their value are less fluctuant.The complete storage of monitoring data bring high information redundancy,which not only waste storage equipment,but also bring great inconvenience to the transmission and analysis real-time data,so data compression processing technology has gradually become an necessary part of real-time data processing.The paper proposes the spectrum data compression method based on pattern recognition by studying and analyzing the regularity of monitoring data in FM broadcasting band,the main research content is as follows:(1)On the basis of the classical K-means algorithm,an modified K-means algorithm is proposed for selecting the initial clustering center and optimizing two parts based on common sense clustering.The initial clustering centers are effectively selected using distance common sense and the number of categories,and clustering optimization is completed through mobile optimization,cross optimization and location optimization based on common sense,it is verified that the modified K-means algorithm has better clustering results in the spectrum data classification.(2)A spectrum data pattern analysis method based on clustering analysis is proposed,study and analyze the regularity of the spectrum data in radio broadcasting bands to classify different category by using the modified K-means algorithm,and extract clustering results as different patterns.The monitoring data under various circumstances are trained to obtain the representative,high recognition and stable pattern sets.(3)A spectrum data compression method based on pattern recognition is proposed,which matches the monitoring data with the patterns of the trained pattern set to record the corresponding pattern labels.And then the original monitoring data is replaced with the pattern set and the corresponding pattern labels,thus realize the compression of the original raw monitoring data.The spectrum data compression method proposed in this paper is used to test the monitoring data of radio broadcasting band,and performance analysis is conducted to verify the feasibility and effectiveness of the method.