Research On Context-aware Signal Processing Technology

With the development of the network, the wildly use of various sensors andmobile devices, ubiquitous computing has been paid more and more attention.Context-aware technology from the so-called ubiquitous computing, is the basis ofnetworking intelligence service, and has become very active research field inelectronic information engineering.Among many related technologies, the frequency spectrum parameters of thesignal (frequency, amplitude and phase) is the most commonly used method to obtainthe location, so the estimation and recovery technology of high-precision signalspectrum parameter become the pursuit of researchers. This work took RFID-basedwireless location and environment-aware applications demand as background, andfocused on the parameter estimation and recovery technology of four kinds of signals,such as short-interval sampling for narrowband low-frequency signals, generalnarrowband signals, broadband non-stationary signals and ultra-wideband signals.The achieved results are list as follows:1．Taking the low frequency estimation problem of short-interval frequencyestimation of narrowband low-frequency signal into account, combined all-phase FFTwith analytic transformation, an frequency estimation algorithm based on analyticall-phase delay-phase difference method is proposed in this paper. The simulationresults show that this method can effectively mitigate the positive and negativespectral interference and spectral leakage. Under the condition of fluctuation CIR in0.4to1without noise, the frequency measurement error can be controlled lower than2%using the proposed method.2．For the deterioration problem of time delay estimation accuracy ofnarrowband signal in the medium and low signal-to-noise ratio area, we propose aphase difference super-resolution time delay estimation method based on all-phasefrequency controlled "point-pass" filter. Verified by simulation, time delay estimationof the proposed method has the apparent better performance compared with thegeneral used barycentric interpolation method under the circumstance of low SNR.3．Considering the waveform restore issues of non-stationary signal, a FRFTtime-frequency filter design method based on time-frequency representation edgeextraction is proposed. Under the premise that the signal and noise are coupled but not overlapped, the proposed method can training the optimal classification line frequencyin the time-frequency plane with fewer data points. The simulation results show thatthe proposed can achieve the goal of filtering the noise and reconstructing the signalwhether or not the signal and noise region can be linearly differentiated. Meanwhile,under the conditions of high sampling rate, the method is to achieve a highersignal-to-noise ratio improvement factor.4．For the problem of backscatter UWB-RFID system structure influencing theTOA estimation accuracy, an LS-TOA estimator is derived in this paper. Roundtripchannel and label-clock not synchronization problems are considered during theestimator modeling. In addition, TOA estimation problem is changed into optimalsolution problem of the objective function with constraints by simplifying theparameter items. The simulation results show that the proposed estimator possessesgood TOAdetection accuracy and ability to resist scattering, even under the conditionof UWB-RAID roundtrip channel.