Improving signal-to-noise ratio of a terahertz signal using a WaveNet-based neural network.

When acquiring a terahertz signal from a time-domain spectroscopy system, the signal is degraded by measurement noise and the information embedded in the signal is distorted. For high-performing terahertz applications, this study proposes a method for enhancing such a noise-degraded terahertz signal using machine learning that is applied to the raw signal after acquisition. The proposed method learns a function that maps the degraded signal to the clean signal using a WaveNet-based neural network that performs multiple layers of dilated convolutions. It also includes learnable pre- and post-processing modules that automatically transform the time domain where the enhancement process operates. When training the neural network, a data augmentation scheme is adopted to tackle the issue of insufficient training data. The comparative evaluation confirms that the proposed method outperforms other baseline neural networks in terms of signal-to-noise ratio. The proposed method also performs significantly better than the averaging of multiple signals, thereby facilitating the procurement of an enhanced signal without increasing the measurement time.

Machine Learning Techniques for THz Imaging and Time-Domain Spectroscopy.

Terahertz imaging and time-domain spectroscopy have been widely used to characterize the properties of test samples in various biomedical and engineering fields. Many of these tasks require the analysis of acquired terahertz signals to extract embedded information, which can be achieved using machine learning. Recently, machine learning techniques have developed rapidly, and many new learning models and learning algorithms have been investigated. Therefore, combined with state-of-the-art machine learning techniques, terahertz applications can be performed with high performance that cannot be achieved using modeling techniques that precede the machine learning era. In this review, we introduce the concept of machine learning and basic machine learning techniques and examine the methods for performance evaluation. We then summarize representative examples of terahertz imaging and time-domain spectroscopy that are conducted using machine learning.

Enhancement of terahertz reflection tomographic imaging by interference cancellation between layers.

This paper proposes a method to enhance terahertz reflection tomographic imaging by interference cancellation between layers. When the gap between layers is small, the signal reflected on the upper layer interferes with that on the lower layer, which degrades the quality of the reconstructed tomographic image in the lower layer. The proposed method estimates the upper-layer reflection signal by system modeling, which is then eliminated from the acquired signal. In this way, it can provide the correct lower-layer reflection signal, thereby improving the quality of the lower-layer tomographic image. The performance of the proposed method was confirmed using computer simulation data and real terahertz reflection data.

Stiffness measurement using terahertz and acoustic waves for biological samples.

A method is proposed to measure sample stiffness using terahertz wave and acoustic stimulation. The stiffness-dependent vibration is measured using terahertz wave (T-ray) during an acoustic stimulation. To quantify the vibration, time of the peak amplitude of the reflected T-ray is measured. In our experiment, the T-ray is asynchronously applied during the period of the acoustic stimulation, and multiple measurements are taken to use the standard deviation and the maximum difference in the peak times to estimate the amplitude of the vibration. Some preliminary results are shown using biological samples.

Pre- and post-processing for tomographic reconstruction of terahertz time-domain spectroscopy.

Reflection-type terahertz tomography is obtained using time-domain spectroscopy. Due to different velocities of the terahertz ray in free space and inside a sample, the tomographic transverse plane is not obtained by a simple reconstruction using time index. A pre-processing method is proposed to compensate for the different velocities of the terahertz ray for tomographic reconstruction. Maximum intensity projection, averaging, and short-time Fourier transform are proposed as post-processing methods along the depth direction for the terahertz tomography. Log-scale display is also suggested for a better visualization. Some experimental results with the pre- and post-processing are demonstrated.

Adaptive compressed sensing for the fast terahertz reflection tomography.

In this paper, an adaptive compressed sensing is proposed in order to enhance the performance of fast tetrahertz reflection tomography. The proposed method first acquires data at random measurement points in the spatial domain, and estimates the regions in each tomographic image where much degradation is expected. Then, it allocates additional measurement points to those regions, so that more data are acquired adaptively at the regions prone to degradation, thereby improving the quality of the reconstructed tomographic images. The proposed method was applied to the T-ray reflection tomography system, and the image quality enhancement by the proposed method, compared to the conventional method, was verified for the same number of measurement points.

Fast terahertz reflection tomography using block-based compressed sensing.

In this paper, a new fast terahertz reflection tomography is proposed using block-based compressed sensing. Since measuring the time-domain signal on two-dimensional grid requires excessive time, reducing measurement time is highly demanding in terahertz tomography. The proposed technique directly reduces the number of sampling points in the spatial domain without modulation or transformation of the signal. Compressed sensing in spatial domain suggests a block-based reconstruction, which substantially reduces computational time without degrading the image quality. An overlap-average method is proposed to remove the block artifact in the block-based compressed sensing. Fast terahertz reflection tomography using the block-based compressed sensing is demonstrated with an integrated circuit and parched anchovy as examples.