To assist with smoking cessation, wearable devices are used to detect the puff (hand-to-mouth gesture) recognition within the smoking activity in a ubiquitous manner. There is a strong assumption that smoking and wearing a smartwatch are usually with the same hand. It will certainly fail to detect smoking gesture with the opposite hand. In this work, we find an interesting phenomenon: smoking can cause a unique pattern of heart rate (HR) which is quite different from other daily activities’ effects. Based on this psychophysiological response, we propose HeartIt, a just-in-time smoking detection solution through measuring the HR by a smartwatch. HeartIt works well for the smoker wearing a smartwatch on either wrist. It can accurately distinguish smoking from other similar hand-to-mouth gestures (e.g., eating, drinking). Moreover, we design an adaptive tracker to trigger the HR sensor once the gesture of lighting a cigarette is detected by low-cost accelerometers. It is robust for different people in various postures and scenarios. Our real-world experiments show that the precision and recall rate of HeartIt reaches 96.7% and 99.8%, respectively.

The influence of non-Independent Identically Distribution (non-IID) data on Federated Learning (FL) has been a serious concern. Clustered Federated Learning (CFL) is an emerging approach for reducing the impact of non-IID data, which employs the client similarity calculated by relevant metrics for clustering. Unfortunately, the existing CFL methods only pursue a single accuracy improvement, but ignore the convergence rate. Additionlly, the designed client selection strategy will affect the clustering results. Finally, traditional semi-supervised learning changes the distribution of data on clients, resulting in higher local costs and undesirable performance. In this paper, we propose a novel CFL method named ASCFL, which selects clients to participate in training and can dynamically adjust the balance between accuracy and convergence speed with datasets consisting of labeled and unlabeled data. To deal with unlabeled data, the prediction labels strategy predicts labels by encoders. The client selection strategy is to improve accuracy and reduce overhead by selecting clients with higher losses participating in the current round. What is more, the similarity-based clustering strategy uses a new indicator to measure the similarity between clients. Experimental results show that ASCFL has certain advantages in model accuracy and convergence speed over the three state-of-the-art methods with two popular datasets.