Study on forearm swing recognition algorithms to drive the underwater power‐assisted device of frogman

Abstract

The frogmen have a great demand for the underwater power-assisted device (UPaD) because of their poor working conditions and complex missions. Although the wearable power-assisted device on land has been diversified, the development of UPaDs is relatively slow. Meanwhile, the most of current small frogman transport equipment is needed to be manipulated by the frogmen with their hands with the aim of changing their moving speed or posture, which increases the difficulty of frogmen executing their tasks. The major contribution of this paper is to develop a method that frees the hands and realizes power assisting at the same time for frogmen, which can be applied to drive the UPaDs of frogmen. Two posture recognition algorithms based on the Human-Computer Interface have been promoted.The algorithms can recognize the forearm swings of the wearer of UPaD, calculate the frequency of the wearer's forearm swing, and adjust the moving speed of the wearer's UPaD according to the wearer's forearm swing frequency. Moreover, we evaluate the performance of two algorithms recognizing forearm swings underwater from four different subjects. Meanwhile, one of the two algorithms is improved and evaluated by experiments from the previous four subjects. The results of accuracy of two algorithms recognizing the wearer's forearm swing are more than 90%. The performance of the speed regulating system of UPaD based on the designed two algorithms is also acceptable.

Abnormal time sequence identification algorithm for power measurement devices based on variational autoencoder and support vector machine

“…Based on Figure 1, it can be seen that the power metering device state analysis model constructed in this article can be mainly divided into two parts. Among them, the reconstruction model, as shown in Figure 1 (a), is composed of a VAE network improved by LSTM [9] , responsible for generating the time series reconstruction data of the power metering device corresponding to the input data. The reconstructed model consists of three parts: inference network, generative network, and random noise [10] .…”

Research on the Application of VR Technology based on Human Posture Recognition in Online Physical Education

“…However, research usually focuses on the technology of teaching platforms or the learning theoretical models. But they pay less attention to the application and optimization of human posture recognition based on VR in physical education, especially in solving the low learning efficiency and low teaching quality in online teaching [1][2][3]. In the traditional offline physical education, there are many challenges that affect teaching efficiency and quality.…”

Research on anomaly identification of electric power metering device based on data driven model

“…With the characteristics of patrol inspection data, the diagnosis of the state of electric energy metering devices is realized, and the corresponding identification results are highly reliable. However, it shows a certain lag in efficiency [4]. In combination with the power consumption status of more and more grid users at this stage, improving the efficiency of the state detection method of electric energy metering equipment is an important basis to ensure the quality of power grid operation and maintenance [5].…”