Sensor-based activity recognition (AR) depends on effective feature representation and classification. However, many recent studies focus on recognition methods, but largely ignore feature representation. Benefitting from the success of Convolutional Neural Networks (CNN) in feature extraction, we propose to improve the feature representation of activities. Specifically, we use a reversed CNN to generate the significant data based on the original features and combine the raw training data with significant data to obtain to enhanced training data. The proposed method can not only train better feature extractors but also help better understand the abstract features of sensor-based activity data. To demonstrate the effectiveness of our proposed method, we conduct comparative experiments with CNN Classifier and CNN-LSTM Classifier on five public datasets, namely the UCIHAR, UniMiB SHAR, OPPORTUNITY, WISDM, and PAMAP2. In addition, we evaluate our proposed method in comparison with traditional methods such as Decision Tree, Multi-layer Perceptron, Extremely randomized trees, Random Forest, and k-Nearest Neighbour on a specific dataset, WISDM. The results show our proposed method consistently outperforms the state-of-the-art methods.
Cancer remains an intractable medical problem. Rapid diagnosis and identification of cancer are critical to differentiate it from nonmalignant diseases. High-throughput biofluid metabolic analysis has potential for cancer diagnosis. Nevertheless, the present metabolite analysis method does not meet the demand for high-throughput screening of diseases. Herein, a high-throughput, cost-effective, and noninvasive urine metabolic profiling method based on TiO2/MXene-assisted laser desorption/ionization mass spectrometry (LDI-MS) is presented for the efficient screening of bladder cancer (BC) and nonmalignant urinary disease. Combined with machine learning, TiO2/MXene-assisted LDI-MS enables high diagnostic accuracy (96.8%) for the classification of patient groups (including 47 BC and 46 ureteral calculus (UC) patients) from healthy controls (113 cases). In addition, BC patients can also be identified from noncancerous UC individuals with an accuracy of 88.3% in the independent test cohort. Furthermore, metabolite variations between BC and UC individuals are investigated based on relative quantification, and related pathways are also discussed. These results suggest that this method, based on urine metabolic patterns, provides a potential tool for rapidly distinguishing urinary diseases and it may pave the way for precision medicine. 相似文献
Metal micropatterns play critical roles in flexible electronics. However, the lack of versatile strategies for micropatterning of diverse metal materials on various thin, flexible or stretchable substrates has limited the rapid development of flexible electronics. Here, a metal micropatterning method by triboelectric spark discharge under atmospheric environment is developed, where a triboelectric nanogenerator (TENG) is employed to precisely and safely control the voltage, current, and frequency of the spark discharges. Micropatterns of metal films like gold, silver, copper, aluminum and platinum are successfully fabricated on substrates of polyimide, polyethylene terephthalate, polyvinyl chloride, polydimethylsiloxane, paper or latex, even on ultrathin substrates (5 μm thick) without damage, where the feature sizes of metal patterns are controllable from 20 μm to 1 mm. Experimental insights into the triboelectric spark discharge behaviors and the pattern feature sizes control are discussed. A straightforward fabrication of metal patterns on the balloon surface or human skin through “handwriting” by a pencil as discharge electrode is realized. Besides metals, extended processibility of conductive materials like carbon nanotubes, graphene, MXene, graphite, carbon fibers, and conductive polymers are also demonstrated. This work proves the possibility of microfabrication by TENG, which is of simplicity and attractiveness for flexible electronics. 相似文献