To get source azimuth from microphone observation angle of view in a complex real environment, this article, on the basis of the analysis of geometric positioning method, established a seven-element microphone array model and proposed a sound source omnidirectional positioning calibration method based on microphone observation angle. By using a seven-element array to invert the position and angle of a sound source, the relative time delay value of a pair of microphones on the vertical axis of the coordinate system is used to determine the elevation angle polarity and realize the omnidirectional sound source positioning. The array parameters, sound velocity, array size, horizontal deflection angle, elevation angle, and sound source are analyzed, and the error method is proposed. The sound source data was measured using the microphone array perspective, and a new Cartesian coordinate system was established based on the observation angle of view for omnidirectional positioning calibration of the sound source. The simulation results show that the position error of the method is about 0.01% and the angle error is about 0.005%, with high calibration accuracy. The actual measurement results show that this method can effectively calibrate the sound source azimuth, the error rate of the source coordinates is around 10%, the horizontal declination angle error is less than 5%, and the elevation angle error is less than 8%. Appropriately increasing the spacing of the array will have a better calibration effect in an actual complex experimental environment.

The measurement of the atmospheric electric field is of great significance for the study of thunderstorm cloud charge models. Traditional electric field meters can only measure the vertical component of the atmospheric electric field, and thus it is difficult to invert the structure of the thunderstorm cloud. A three-dimensional atmospheric electric field meter was developed to simultaneously measure the horizontal and vertical components of the atmospheric electric field in this paper. The effective measurement linearly relates the measured electric field to the induced voltage, and the nonlinear equations of the three-dimensional atmospheric electric field and the thunderstorm cloud-charging model parameters were derived. The particle swarm optimization algorithm (PSO) and the three-dimensional atmospheric electric field were used to invert the thunderstorm clouds. Experimental observations of the three-dimensional electric field in a cloud during a thunderstorm were analyzed. Combined with the typical charged structure model, parameters such as the charge and relative distance of the thunderstorm cloud were determined. The results showed that the value of the inversion fitness function reached 0.7288, and the charge structure was even. The measurement of the three-dimensional atmospheric electric field provides a new means of observation for the study of atmospheric electricity.

No abstract

In order to obtain the position of thunderstorm cloud in real time and make it possible to track the thunderstorm cloud motion, a method is proposed for tracking the moving path of thunderstorm cloud, with the aid of the three-dimensional atmospheric electric field apparatus (AEFA). According to the method of images, we establish a spatial model for tracking the moving path. Based on the model, we define the dynamic parameters of thunderstorm cloud position. Subsequently, to realize the moving path tracking of thunderstorm cloud, its coordinates are associated with the time points. Besides, we use the relationship between electric field component measurement error, horizontal angle, elevation angle, and the tracking accuracy to analyze the tracking performance. Finally, a fusion system combining an electric field measurement unit, electric field calibration unit, and permittivity measurement unit is designed to meet the actual needs. The results show that the method can accurately track the thunderstorm cloud moving path and has a better effect. In addition, the method can also be combined with a radar map, thus better predicting the development of the thunderstorm cloud.

Federated Learning (FL) enables the multiple participating devices to collaboratively contribute to a global neural network model while keeping the training data locally. Unlike the centralized training setting, the non-IID, imbalanced (statistical heterogeneity) and distribution shifted training data of FL is distributed in the federated network, which will increase the divergences between the local models and the global model, further degrading performance. In this paper, we propose a flexible clustered federated learning (CFL) framework named FlexCFL, in which we 1) group the training of clients based on the similarities between the clients' optimization directions for lower training divergence; 2) implement an efficient newcomer device cold start mechanism for framework scalability and practicality; 3) flexibly migrate clients to meet the challenge of client-level data distribution shift. FlexCFL can achieve improvements by dividing joint optimization into groups of sub-optimization and can strike a balance between accuracy and communication efficiency in the distribution shift environment. The convergence and complexity are analyzed to demonstrate the efficiency of FlexCFL. We also evaluate FlexCFL on several open datasets and made comparisons with related CFL frameworks. The results show that FlexCFL can significantly improve absolute test accuracy by +10.6% on FEMNIST compared to FedAvg, +3.5% on FashionMNIST compared to FedProx, +8.4% on MNIST compared to FeSEM. The experiment results show that FlexCFL is also communication efficient in the distribution shift environment.