Fig. 1. Original spectra of seven weight loss drugs

Fig. 2. Characteristic peak interval of seven weight loss drugs

Fig. 3. Flow chart of random forest model

Fig. 4. Recognition accuracy of PSO-LSSVM under different feature fusion methods for seven weight loss drugs. (a) Original spectra; (b) feature fusion spectra after Hilbert transform; (c) feature fusion spectra after SNV+DT transform; (d) feature fusion spectra after FFT low-pass filter transform; (e) feature fusion spectra after Butterworth low-pass filter transform

Fig. 5. Confusion matrix of random forest model under different feature fusion methods for seven weight loss drugs. (a) Original spectra; (b) Hilbert transform; (c) SNV+DT transform; (d) FFT low-pass filter transform; (e) Butterworth low-pass filter transform

Fig. 6. Accuracy of random forest model under different feature fusion methods for seven weight loss drugs. (a) Original spectra; (b) Hilbert transform; (c) SNV+DT transform; (d) FFT low-pass filter transform; (e) Butterworth low-pass filter transform

Fig. 7. Classification and recognition accuracy of PSO-LSSVM and RF. (a) PSO-LSSVM; (b) RF

Fig. 8. THz time-domain spectrum of sample S-1

Fig. 9. THz time-domain spectrum of sample S-2

Fig. 10. THz time-domain spectrum of sample S-3

Table 1. Statistics of weight loss drug samples

Table 2. Basic information of equipment and parameters

Table 3. Classification result of three samples to be judged