Genetic-Algorithm-Based Optimization for Terahertz Time-Domain Adaptive Sampling

摘要

We propose a genetic algorithm (GA)-based method to improve the sampling efficiency in terahertz time-domain spectroscopy (THz-TDS). For a typical time-domain THz signal, most information are contained in a short region of the pulse, which needs to be densely sampled, while the other regions fluctuating around zero can be represented by fewer points. Based on this clustering feature of the THz signal, we can use much fewer sampling points and optimize the distribution by using a GA to achieve an accurate scanning in less time. Both reflection and transmission measurements were conducted to experimentally verify the performance. The measurement results show that the sampling time can be greatly reduced while maintaining a very high accuracy in both time and frequency domains compared with a high-resolution step scan. This method significantly improves the measurement efficiency. It can be easily adapted to most THz-TDS systems equipped with a mechanical delay stage for fast detection and THz imaging.