Federated Learning (FL) enables smart devices to collaboratively train Machine Learning (ML) models in a distributed manner without sharing their private data with a central server. However, the disparity between the communication and computation capabilities, and the heterogeneity of local datasets of smart devices degrades the performance of FL in terms of latency and accuracy. To mitigate this effect, we address the problems of device selection and resource allocation in an indoor environment where multiple smart devices participate in the FL process. To further reduce the communication latency, we use Visible Light Communication (VLC) for the downlink transmission while a Radio Frequency (RF) access point supports the uplink transmission in the proposed system. Accordingly, we formulate a multi-objective optimization problem for joint device selection and resource allocation in a hybrid VLC/RF system. Then, using the weight methods, the problem is converted to a single-objective optimization which is solved by incrementally selecting devices in each iteration. The embedded device selection scheme in the proposed algorithm is based on the significance of candidate devices' local gradients and their alignment with the global tendency in order to intelligently prioritize the candidates in the training procedure. Simulation results show that the joint device selection and resource allocation scheme improves the accuracy of the ML model and reduces the average delay in presence of both system and data heterogeneity. Additionally, the proposed hybrid VLC/RF system decreases the latency of the FL process in the downlink mode compared to conventional RF systems.INDEX TERMS Visible light communication, federated learning, device selection, resource allocation, internet of things.

In recent years, Deep Neural Networks (DNNs) have been widely used for Human Gesture Recognition (HGR) based on the information obtained from inertial sensors, such as accelerometers and gyroscopes, available on smart Internet of Things (IoT) devices. Most of the recent works on HGR using motion data rely on gathering a dataset, that faces two major challenges:a) the datasets are originally stored on the smart devices at the end-users, and gathering them in one place is not feasible due to communication limitations, andb) clients are reluctant to share their private data with a central server due to privacy concerns. In this paper, we address these issues and propose a privacy-preserving framework based on Federated Learning (FL) for HGR using motion data, called Motion-based Federated Learning Gesture Recognition (MoFLeuR). Furthermore, we consider different types of data heterogeneity which have destructive effects on the performance of the global model. Accordingly, we propose a communication and computation-efficient client selection method that chooses the clients to mitigate the impact of data heterogeneity in the training process. In the proposed framework, clients are not requested to share sensitive information about their local datasets with the edge server in the FL process. Simulation results show that the proposed MoFLeuR algorithm improves the performance of the global model in the presence of different degrees of data heterogeneity, and it outperforms the baseline algorithms in terms of different metrics, namely accuracy, convergence speed, and communication and computation efficiency.