| In the forensic pathology investigation of the traffic accidents all over the world, drunken driving is an important reason leading to traffic accidents. Detection and prevention of drunken driving draw more and more attentions in the world. At present, the mature methods of alcohol detection are breath alcohol detection and blood test. The blood test is invasive detection method and the breath alcohol detection is inaccuracy. On the contrary, near-infrared spectroscopy (NIRS) as a rapid, accurate, non-destructive detection method, compared with existing technology, has superior advantages. Near-infrared spectroscopy has been widely used to detect the human body substances, such as the detection of glucose, oxygen content, we can say that the technology has entered a mature stage of development, however, there has not been a method to detect the blood alcohol content of the drivers at home and abroad.The purpose of this paper is to realize the noninvasive detection of the blood alcohol content of the drivers based on the near-infrared spectroscopy. The paper studies the near-infrared diffuse reflectance spectroscopy of blood alcohol content with the methods of wavelet analysis and partial least squares method and studies the law of the establishment and absorption of blood alcohol in the human body. A quantitative model based on the partial least squares method to detect the alcohol content is established and the stability and the precision are tested.The near-infrared spectroscopy signals in vivo have many noises, especially when the test is collected directly on the surface of the skin, which can not be directly used to establish the model. The pretreatment of the near-infrared spectroscopy we use in this paper is wavelet analysis. The paper compares the differences of the default de-noising method in matlab, Birge-Massart de-noising method and maximum and minimum de-noising method for in vitro and in vivo alcohol near-infrared diffuse reflectance spectroscopy under the conditions of the soft threshold and hard threshold. The results are evaluated according to noise ratio and the root mean square error of different methods. As a result, the default de-noising method in matlab under the conditions of the hard threshold is chosen as the method to de-noise spectral data obtained. From the results of de-noising, we determine the characteristics of the spectral wavelength of the alcohol ranges from2200nm to2400nm. With the spectral data based on wavelet analysis, partial least squares (PLS) method was utilized to establish the best mathematic calibration model that for describing the relationship between the near-infrared information and each component. The optimal principal components is determined by leave-one-out cross validation. The paper adopts the correlation coefficient (R), root mean square error (RMSEC) as the evaluation parameters of the calibration model, the prediction mean square error (RMSEP) and the average relative error (MREP) to evaluate the results of the prediction of unknown samples. And we evaluate the repeatability of the model, which turns out to be a good result. In other words, the model can be used to verify the alcohol concentration detection.At the end, According to different people, different alcohol consumption, individual differences, the paper listed a detailed classification and the law of the metabolism of the blood alcohol is studied under different situations. Studies have shown that the metabolism of alcohol in the blood generally presents a first increases, then a peak, and then comes to a decrease. Fifty minutes to seventy minutes after the human drinking alcohol, the blood alcohol content reaches its maximum. The metabolism of alcohol in the body has a great relationship with alcohol consumption, fullness or fasting, personal physical fitness. Fasting, excessive alcohol consumption and the lack of aldehyde dehydrogenase2in vivo contribute to drunken. Alcohol residues stay in the human body longer than normal. |
PDF Full Text Request