Pedestrian positioning with wearable devices is a significant application of attitude tracking. It tracks the attitude (with heading angle being the most important part) of the device in real time and provides positioning services for users based on the information of step length provided by Pedestrian Dead-Reckon (PDR), which is a cheap and efficient positioning method at present. However, amid a train of positioning methods, the joint estimate of tracking is given by a train of methods based on the direction of gravity and the earths magnetic field direction. Considering the measurement of gravity that the gravity accelerometer is exposed to heavy noise due to the complex movement of human body during walking with uniform swing arm posture and forward speed, this paper proposed a novel estimate method based on the Kalman filter with multi-state constraints and the usage of low-cost sensors, which fulfills the estimation with the sequential observation of magnetic field. Compared with other related work, this method proposed in this paper eliminates the dependence on gravity direction, avoiding the influence of heavy noise caused by additional linear acceleration in motion state, and reduces the influence of insufficient observation when using magnetic field observation alone. The performance of the proposed method is evaluated by real-world experimentation results.

This paper proposes a new algorithm for adaptive deep image compression (DIC) that can compress images for different purposes or contexts at different rates. The algorithm can compress images with semantic awareness, which means classification-related semantic features are better protected in lossy image compression. It builds on the existing conditional encoder-based DIC method and adds two features: a model-based rate-distortion-classification-perception (RDCP) framework to control the trade-off between rate and performance for different contexts, and a mechanism to generate coding conditions based on image complexity and semantic importance. The algorithm outperforms the QMAP2021 benchmark on the ImageNet dataset. Over the tested rate range, it improves the classification accuracy by 11% and the perceptual quality by 12.4%, 32%, and 1.3% on average for NIQE, LPIPS, and FSIM metrics, respectively.

No abstract