WiFi-based human activity recognition (HAR) holds significant promise for ubiquitous sensing in smart environments. A critical challenge lies in enabling systems to dynamically adapt to evolving scenarios, learning new activities without catastrophic forgetting of prior knowledge, while adhering to the stringent computational constraints of edge devices. Current approaches struggle to reconcile these requirements due to prohibitive storage demands for retaining historical data and inefficient parameter utilization. We propose WECAR, an end-edge collaborative inference and training framework for WiFi-based continuous HAR, which decouples computational workloads to overcome these limitations. In this framework, edge devices handle model training, lightweight optimization, and updates, while end devices perform efficient inference. WECAR introduces two key innovations, i.e., dynamic continual learning with parameter efficiency and hierarchical distillation for end deployment. For the former, we propose a transformer-based architecture enhanced by task-specific dynamic model expansion and stability-aware selective retraining. For the latter, we propose a dual-phase distillation mechanism that includes multi-head self-attention relation distillation and prefix relation distillation. We implement WECAR based on heterogeneous hardware using Jetson Nano as edge devices and the ESP32 as end devices, respectively. Our experiments across three public WiFi datasets reveal that WECAR not only outperforms several state-of-the-art methods in performance and parameter efficiency, but also achieves a substantial reduction in the model's parameter count post-optimization without sacrificing accuracy. This validates its practicality for resource-constrained environments.

WiFi-based human activity recognition (HAR) holds significant application potential across various fields. To handle dynamic environments where new activities are continuously introduced, WiFi-based HAR systems must adapt by learning new concepts without forgetting previously learned ones. Furthermore, retaining knowledge from old activities by storing historical exemplar is impractical for WiFi-based HAR due to privacy concerns and limited storage capacity of edge devices. In this work, we propose ConSense, a lightweight and fast-adapted exemplar-free class incremental learning framework for WiFi-based HAR. The framework leverages the transformer architecture and involves dynamic model expansion and selective retraining to preserve previously learned knowledge while integrating new information. Specifically, during incremental sessions, small-scale trainable parameters that are trained specifically on the data of each task are added in the multi-head self-attention layer. In addition, a selective retraining strategy that dynamically adjusts the weights in multilayer perceptron based on the performance stability of neurons across tasks is used. Rather than training the entire model, the proposed strategies of dynamic model expansion and selective retraining reduce the overall computational load while balancing stability on previous tasks and plasticity on new tasks. Evaluation results on three public WiFi datasets demonstrate that ConSense not only outperforms several competitive approaches but also requires fewer parameters, highlighting its practical utility in class-incremental scenarios for HAR.

HAR is a technique that aims at predicting participants' activities from the lowlevel sensor inputs. In recent years, the use of wearable devices such as smart wristbands and smart phones has significantly facilitated HAR research due to their small sizes and low power consumption [5] as well as their capability of real-time information capturing [6]. Traditional machine learning (ML) methods such as support vector machines and decision trees have significantly promoted the development of HAR studies during the past few decades. However, limitations such as heavy reliance on human domain knowledge [7] and shallow feature extraction impede these ML methods from providing satisfactory results in most daily HAR tasks [8]. More recently, deep learning (DL) methods have been continuously providing marvelous performance in many areas such as computer vision and natural language processing [9]. The capability to automatically extract high-level features makes DL methods largely alleviated from the drawbacks of conventional ML methods. There have been many DLbased HAR works proposed in recent years [10, 11, 12, 13]. However, the training time and the amount of data required for DL systems are always much larger than that of traditional ML systems, and the large time and labor costs makes it difficult to build a large-scale labeled human activity dataset with high quality. Therefore, it is necessary to find a way to alleviate these problems of DLbased HAR.

Self-taught learning is a technique that uses a large number of unlabeled data as source samples to improve the task performance on target samples. Compared with other transfer learning techniques, self-taught learning can be applied to a broader set of scenarios due to the loose restrictions on source data. However, knowledge transferred from source samples that are not sufficiently related to the target domain may negatively influence the target learner, which is referred to as negative transfer. In this paper, we propose a metric for the relevance between a source sample and target samples. To be more specific, both source and target samples are reconstructed through a single-layer autoencoder with a linear relationship between source samples and target samples simultaneously enforced. An l_{2,1}-norm sparsity constraint is imposed on the transformation matrix to identify source samples relevant to the target domain. Source domain samples that are deemed relevant are assigned pseudo-labels reflecting their relevance to target domain samples, and are combined with target samples in order to provide an expanded training set for classifier training. Local data structures are also preserved during source sample selection through spectral graph analysis. Promising results in extensive experiments show the advantages of the proposed approach.

Feature selection is a dimensionality reduction technique that selects a subset of representative features from high-dimensional data by eliminating irrelevant and redundant features. Recently, feature selection combined with sparse learning has attracted significant attention due to its outstanding performance compared with traditional feature selection methods that ignores correlation between features. These works first map data onto a low-dimensional subspace and then select features by posing a sparsity constraint on the transformation matrix. However, they are restricted by design to linear data transformation, a potential drawback given that the underlying correlation structures of data are often non-linear. To leverage a more sophisticated embedding, we propose an autoencoder-based unsupervised feature selection approach that leverages a single-layer autoencoder for a joint framework of feature selection and manifold learning. More specifically, we enforce column sparsity on the weight matrix connecting the input layer and the hidden layer, as in previous work. Additionally, we include spectral graph analysis on the projected data into the learning process to achieve local data geometry preservation from the original data space to the low-dimensional feature space. Extensive experiments are conducted on image, audio, text, and biological data. The promising experimental results validate the superiority of the proposed method.