关键词: 柔性应变传感器, 机器学习, 信号处理, 智能纺织品, 智能材料

Abstract:

Significance Because of the rapid progress and growth of smart materials and smart textiles, increasing attention hasbeen focused on the research, development, and optimization of flexible strain sensors. Flexible strain sensors for smart textiles are capable of detecting the precise motion trajectory of the human body, mechanical-acoustic characteristics, and information on various physiological indicators. With the continuous optimization of the performance of flexible strain sensors, the flexible sensor devices need to achieve the acquisition and analysis of high-dimensional and high-frequency complex superimposed signals in very complex application environments, which in turn puts forward higher requirements for data processing algorithms. The implementation of machine learning, a more advanced method, has significantly contributed to the improvement in the overall performance of the flexible strain sensing system. This paper presents a systematic review of the research progress of flexible strain sensors based on smart textiles combined with machine learning. The goal of the review is to understand and broaden the application of machine learning in the field of flexible strain sensors.

Progress This paper firstly made an in-depth analysis of the fundamental structure and previous research on a variety of conventional flexible strain sensors such as piezoresistive, piezoelectric, capacitive, optical, magnetic, and triboelectric. In addition, this paper introduced the workflow of machine learning, which can be divided into the following four main steps: data preprocessing, machine learning and model training, model evaluation, and prediction of new data. According to the learning method, machine learning can be classified into supervised learning, unsupervised learning, reinforcement learning, and a mixture of the above three types. This paper then paper provided a detailed description of the information processing process of flexible strain sensors based on machine learning, as well as summarized the advantages and disadvantages of some typical machine learning algorithms for time-frequency analysis, dimensionality reduction, and classification. Furthermore, this paper analyzed the most recent research on flexible strain sensors based on smart textiles combined with machine learning in the fields of healthcare, life assistance, communication and exchange, as well as teaching and entertainment, which placed a significant amount of emphasis on the benefits that can be gained from utilizing machine learning in flexible strain sensors. In the field of healthcare, flexible strain sensing systems can continuously track various mechanical and acoustic features of the human body by combining with specific machine learning algorithms, which can help users to understand their own health status in real time, and thus achieve the purpose of health monitoring. Secondly, in the field of life assistance, the large amount of information provided by the machine learning-based strain sensing system can help in the design of bionic hearing, touch, and prosthetic manipulator, which can greatly improve the convenience of life for the disabled and the blind. Moreover, free-life monitoring by flexible strain sensing systems has the potential advantage of accurately detecting and measuring clinically relevant features, including fall risk and abnormal gait, so that abnormal movement symptoms of the elderly can be detected in a timely manner, which can ensure the safety of the elderly's life to a considerable extent. In the field of communication and exchange, the application of flexible strain sensors based on machine learning can improve the recognition performance of various features, such as sign language recognition, micro-expression detection, and perceptual interaction, thus facilitating human-to-human communication. In addition, the strain sensing system combined with specific machine learning algorithms enriches the application of smart textiles in teaching and entertainment scenarios, which improves the teaching efficiency and enhances the fun of teaching at the same time, and the application in gaming and entertainment greatly enriches people's lives.

Conclusion and Prospect Flexible strain sensors have excellent characteristics, such as high sensitivity, high resolution, and good elasticity. With the help of new sensor structures, new sensitive materials, and cutting-edge machine learning algorithms, smart textiles have been of great value in a variety of different fields. However, in the context of material intelligence, the research on flexible strain sensors based on smart textiles is still in its infancy and still faces many challenges, such as the fact that researchers have carried out little research on the optimal design of flexible strain sensor arrays, that it is difficult to simulate real human touch with flexible strain sensors designed according to existing technologies, and that the process of human pose recognition with flexible strain sensor systems can easily cause confusion in the recognition system. In a word, there is no doubt that machine learning has evolved into a valuable tool in the realm of smart wearables. It is believed that in the near future, with the continuous development of computer science and computing methods, machine learning will play a huge application value in various aspects such as the research and development of smart textile materials, process improvement, device performance evaluation, signal transmission, data processing, etc., and will further promote the intelligent development of the whole material field.

Key words: flexible strain sensor, machine learning, signal processing, smart textile, smart material