Novel Methods For Spectral Data Quality Evaluation And Frequency Domain Extraction Of Dynamic Spectrum

In recent years,noninvasive blood component detection technology has been the cutting-edge research hot spot in the field of bio-medical detection.Optical method makes the measurement of blood components safe,noninvasive,efficient and fast.Among them,the dynamic spectrum method,a method of great potential so far used for the detection,has eliminated the impact of individual differences and measurement environment in principle.In the process of dynamic spectroscopy analysis,satisfying modeling prediction results can only be made by the spectral data of high quality and dynamic spectrum extraction method of favorable efficiency.Therefore,the quality evaluation and selection of spectral data as well as the extraction of high quality dynamic spectrum(DS)from the photoelectric volume pulse wave(PPG)signals play an important part in the noninvasive blood component testing of the signal dynamic spectrum method.In this paper,based on the basic principle and method of dynamic spectrum,the following research work has mainly been carried out around the above arguments.Aiming at the problem in the study of dynamic spectroscopy,that the evaluation method of the data quality is single and the accuracy is not high,a new method of spectral data quality evaluation and selection based on Fundamental Frequency Stability Coefficient(FFSC)is proposed in this paper,and its feasibility has been verified by the simulation and clinical samples experiment.In the clinical samples experiment,by the method of grouping contrast,the samples were selected and their dynamic spectrum data were extracted,then the partial least square method is adopted to establish the model about the dynamic spectral data of the subject and the biochemical value of hemoglobin.Compared with the unfiltered random sample group,the prediction accuracy of the preferred sample group is improved by 12.19% through the fundamental frequency stability analysis,which greatly improved the prediction accuracy of the model.On account of the shortcomings of the traditional dynamic spectral frequency domain extraction method,that it cannot reflect the characteristics of the local area of the time domain signal and is prone to error,an improved frequency domain extraction method is proposed.The new method adopted Short-time Fourier transform(DSFT)combined with statistical processing method,which can effectively eliminate the singular value in the photoelectric volume pulse wave(PPG)signals and maximize the use of the sampled data,thus significantly improve the quality of the acquired dynamic spectrum(DS).This study was based on the clinical data of the subjects,and the hemoglobin content was used as the analysis object.The experimental group employed the dynamic spectral data obtained by the dynamic spectral extraction method based on the Short-time Fourier transform,while the control group employed the data obtained by the traditional dynamic spectral frequency domain extraction method.And both groups respectively established a model for analysis.Compared with the control group,the model predictive correlation coefficient of the experimental group increased by 0.19,and the root mean square error was reduced by 3.3g/L.The experimental results show that the method can improve the stability and precision of the model in the upper process of modeling prediction for hemoglobin concentration and improve the effective utilization rate of PPG signal.The research in this paper has laid a foundation for subsequent clinical analysis,and also provides reference and new ideas for the application of spectral analysis in other fields of signal processing.