Studies On The Filter Methods Of Capillary Electrophoresis Signals Based On Wavelet Transform

Capillary electrophoresis (CE) has become an important analytical method in pharmaceutical analysis field. Due to the basic theory, uncertain environmental disturbance and various electrophoresis conditions, noise and real CE signals often coexist in the CE profile, which will greatly affect the accuracy and reproducibility of analytical results. Therefore, the filter of CE signals is significant to the accuracy and validity of analytical results. In the present thesis, three de-noising methods were proposed to deal with different noise conditions:1) The reported de-noising methods only deal with off-line CE signals and can not be applied to real-time monitoring. To solve this problem, the online wavelet transform with fixed time window was used and the selection of the best length of time window was investigated. The simulated signals and the CE signals of Cordyceps sinens served as examples to test this method. It was indicated that this method could significantly improve the signal-to-noise ratio (SNR) and efficiently filter the noise peaks, while the peak position and height remained unchanged.2) Obvious error of peak value may appear if the signals with small full width at half-maximum (FWHM) were filtered with traditional smoothing and de-noising approaches. Therefore, the second generation wavelets (SGWT) algorithm based on interpolating subdivision was proposed in this thesis to minimize the possible lost of information in the signal processing. The CE signals of Saponins standard solution were de-noised by this method and the SNR was significantly improved. Compared with the smoothing method, the relative error of peak height and FWHM was decreased by 15.1% and 16.7%, respectively. Furthermore, the limit of detection (LOM) and the lowest concentration of linear range were decreased to only 27.0% and 27.8% of the original, respectively.3) For analyzer lacking of knowledge of signal processing, it is difficult to select appropriate filter parameters. Therefore, an adaptive wavelet transform de-noising method was proposed with the aim to detect the signal features automatically bystudying the correlation among samples. The simulation results showed that this method can be successfully applied in dealing with different kinds of signals with various SNRs and FWHMs. Application of the proposed method to CE signals of Saponins standard solution also exhibited significantly improvement with LOM and the lowest concentration of linear range decreased to 14.1% and 25.0% of the original, respectively.