On Signal Detection Using Adaptive Cone-kernel Time-Frequency Representation

• Main Authors: Liu, Jun-Cheng, 劉俊成
• Other Authors: Chang Shun-Hsyung
• Format: Others
• Language: zh-TW
• Published: 1997
• Online Access: http://ndltd.ncl.edu.tw/handle/82353292812315604274

碩士 === 國立海洋大學 === 電機工程學系 === 85 === For years, flocks of engineering researchers have been working forthe accuracy enhancement of time-frequency signal detection due to itsextensive applications in the fields of RADAR, SONAR, Seismology andunderwater acoustics. Most of the current high resolutionaltime-frequency methods are based on the structure of the popular Cohentime-frequency distribution function, but the resolution achieved isstill requiring improvements. In this thesis, we propose the modifiedcone time- frequency function which is also named as MACK(modified adaptive cone kernel). The researchcarried will concentrate on the discussion of the following threeclass: (1) Proposing a brand new high-resolutinal time-frequencydistribution function according to the excellent features oftime- frequency distribution function. (2) Analyzing the features ofmulti time-frequency distribution function and applying oftime-frequency distribution function to detect the accuracy of theinstantaneous time-frequency rate multi-nonstationarysignal. (3) Applying the time-frequency analysis to array signalprocessing.Although, the time- frequency distribution functions can accuratelytrace the instantaneous frequency contained by nonstaitionary signal,not can all of them analyze the multi nonstationary signal like Wignerdistribution function. For the later case the nonstationary signal hashigher cross terms and vulnerable. In this thesis, we apply thebenficials of cone, exponential and Gaussian kernel to enhanceresolution and further constructiong a new kernel distributionfunction furthermore,the adaptively nature of the kernel distributionfunction is also well utilized in order to arrive at aoptimun-paramentered time-frequency distribution function. Here, weincorporate the newly-developed time-frequency distribution functionwith the technique of high resolutional array signal of processing.Bysuch a combination, the cross terms are suppressed and the problemfrequency changes which are caused by the phasing and timing delays ofarray signal is also manipulated. To verify the validity oftime-frequency is merging with the processing rules of arraysignal. The physical properties can be read from the nuance of thefrequency.The distribution function developed in this thesis is broadlyapplied to the parameter determination of DOA in the array signalprocessing no matter the far-field, the near-field, the single orunder the multi-signal environments. Besides, we are alsoauthencifying the general formula and extracting the data oftime-varying spectrum by using the motion sensor and DOA distance.Most importanty, our method eliminate the redundantcomputational load of multi-dimensional Fourier Projection and thecomplexity of feature analysis and spectrum scanning.