基于离散极大值法的太赫兹光谱特征吸收峰提取方法

太赫兹波在电磁波谱中介于微波和红外辐射之间，具有指纹特性、安全无损、强穿透性等特点，因此太赫兹光谱技术在药品成分和组成检测领域具有广泛应用价值。针对高纯度物质识别研究中存在部分弱吸收峰不易识别，以及混合物的太赫兹光谱中吸收峰强度降低而导致吸收峰位信息模糊化的问题，提出了一种基于离散极大值法的光谱吸收峰位识别方法，即伴随拐点法。伴随拐点法首先利用目标检测物太赫兹吸收系数谱图的一阶和二阶导数确定吸收峰位的伴随拐点和基线谱，其次将原始吸收光谱与基线谱进行差分运算得到差谱，最后根据离散极大值法确定吸收峰位，从而实现特征吸收峰的识别。为验证伴随拐点法的有效性，采用伴随拐点法对四种硝基呋喃类样品光谱进行吸收峰提取，并将吸收峰位识别结果与仿真结果进行比较。实验结果证明，伴随拐点法能有效识别目标检测物的强吸收峰和弱吸收峰。该方法不仅在含峰目标物的太赫兹特征吸收峰识别问题中具有广泛的应用前景，还适用于其他光谱的谱峰峰位检测。

Research on Discrimination Method of Absorption Peak in Terahertz Regime

Terahertz radiation bridges the gap between the microwave and optical regimes. It has unique properties such as fingerprint characteristics, non-destructive testing and transparency to various materials, which makes terahertz waves have significant scientific and technological potential in drug testing applications. Terahertz time-domain spectroscopy plays an important role in identifying target drugs containing absorption peaks. It can be used to discriminate specific molecules contained in drugs or the changes of components in samples, as many molecules have characteristic absorption peaks in the terahertz regime. Thus, to solve the problems of identifying weak absorption peaks of low content targets in the mixture, in this paper, the adjoint inflection point (AIP) method based on the discrete local maximum (DLM) method is proposed for identifying the characteristic absorption peaks in terahertz regime for the effective identification of the low content target. Firstly, the adjoint inflection points of potential absorption peaks are obtained using the first and second derivative of the terahertz absorption coefficient spectrum. Secondly, the difference spectrum is calculated by performing the operation between the original absorption spectrum and the baseline spectrum. At last, the absorption peak positions are determined by using the DLM method along the difference spectrum. Also, this AIP method is applied to the absorption peak extraction of four nitrofurans sample spectra. The result is compared with the peak positions determined by DLM, and the peak positions are also compared with the peaks calculated by the density functional theory. The better performance of the recognition capacity of the AIP method is observed and verified, especially for weak absorption peaks. This method suggests that it has profound application potential in spectroscopic analyses and even in determining curve peaks in various applications.