Federated Self-Training for Data-Efficient Audio Recognition | IEEE Conference Publication | IEEE Xplore