Using the data collected by the sensors embedded in smartphones for human activity recognition has become an important research direction. In order to overcome the problem of handcrafted feature extraction in traditional machine learning, we propose an attention-based model which combines convolutional neural network (CNN) and bidirectional gated recurrent unit (BiGRU) for human activity recognition. This model can extract local features of the original data by CNN, and construct the extracted features into a time series as the input of BiGRU, the BiGRU can capture the temporal dependencies and extract the features from the input, by introducing the attention mechanism, the influence of key input can be strengthened. The features are classified by the Softmax function. The recognition accuracy rate on the Wireless Sensor Data Mining (WISDM) dataset reaches 98.5%, which is improved compared with the state-of-the-art methods.

The time maintenance accuracy of the navigation constellation determines the user positioning and timing performance. Especially in autonomous operation scenarios, the performance of navigation constellation maintenance time directly affects the duration of constellation autonomous navigation. Among them, the frequency stability of the atomic clock onboard the navigation satellite is a key factor. In order to further improve the stability of the navigation constellation time-frequency system, combined with the development of high-precision inter-satellite link measurement technology, the idea of constructing constellation-level synthetic atomic time has gradually become the development trend of major GNSS systems. This paper gives a navigation constellation time scale generation framework, and designs an improved Kalman plus weights (KPW) time scale algorithm and time-frequency steer algorithm that integrates genetic algorithms. Finally, a 30-day autonomous timekeeping simulation was carried out using the GPS precision clock data provided by CODE, when the sampling interval is 300 s, the Allan deviation of the output time scale is 5.73 × 10⁻¹⁴, a 71% improvement compared with the traditional KPW time scale algorithm; when the sampling interval is 1 day, the Allan deviation is 9.17 × 10⁻¹⁵; when the sampling interval is 1 × 10⁶ s, the Allan deviation is 8.87 × 10⁻¹⁶, a 94% improvement compared with the traditional KPW time scale algorithm. The constellation-level high-precision time scale generation technology proposed in this paper can significantly improve the stability performance of navigation constellation autonomous timekeeping.

China is planning to launch a relay satellite in the coming years with a payload for the Earth-Moon Very-Long-Baseline Interferometer (VLBI) observation research and scientific applications, known as the Lunar Orbit VLBI Experiment (LOVEX) system. This system will perform the lunar orbit VLBI experiments for in the world, and will improve the accuracy of planetary spacecraft orbit determination and the scientific observation capabilities of astrophysics and astrometry. Prior to launch, key technologies of the system must be verified, so we have established a LOVEX ground validation system with similar performance to the LOVEX system. The LOVEX ground validation system consists of a 4.5-meter ring-focus antenna, X-band receiver, frequency converter and digital backend, hydrogen maser, and monitoring and control software. Using this system, we have successfully conducted VLBI differential observation experiments with the Shanghai 65m radio telescope and Beijing 50m radio telescope. The experimental results show that, after calibration with NRAO150, the standard deviations of delay error of Tianwen-1 is within 0.5 ns, consistent with the theoretical timing accuracy, verifying the feasibility of the LOVEX time delay calibration.