Xingkai Zheng

Gait recognition of the human skeleton based on Dynamic Vision Sensor I took advantage of the high-speed dynamic capture of DVS equipment to improve the accuracy and speed of gait recognition. Then I cumulatively mapped DVS point cloud data into video data and removed video noise through image processing algorithms. After that, I designed 5 kinds of local artificial skeleton features extracted from video data and used clustering to reduce the dimensions. Finally, I built a deep learning (DL) residual network to use local artificial skeleton features for gait recognition.

Search Engine Group In general, I optimized search engine algorithms serving 300 million customers to purchase various goods in PDD App. I Rebuilt the search AI model training data stream to shorten the training time of 1-2 hours and improve the model accuracy. Besides, I created a conversion delay model with 82% accuracy to prevent samples from entering the search model at the wrong time with my colleagues.

I systematically learned the relevant knowledge of Computer Science and completed a series of course projects using C++, JS, Python, SQL, and other languages. Also, I did some computer vision research, specifically gait recognition, following a professor in my department.