THIS README IS GENERATED BY YANDEX. FOR CN README, SEE README_CN.md.
The code of the open source part of this project follows the MIT agreement and can be used, forwarded, and used for commercial purposes by anyone. Since README does not support video playback, you need to navigate the corresponding project location by yourself to view it.：
The existing wireless perception research mainly focuses on WiFi, while the perception research based on 5GNR has very few examples. As far as I understand, there is only one IEEE article on the use of 5GNR for human activity perception currently published in authoritative journals, and the relevant code and implementation are not open source.
-Display of fall detectionvideos\FallDetect.mp4. In this demo, the content displayed on the right side of the visualization platform of SRS-Front is the status indication of fall/non-fall, the green mark is not obtained, and the red alarm symbol is a fall. The demo shows a number ofAngles and different ways of falling can be accurately identified.
In order to make full use of the flexible and multi-antenna characteristics of 5GNR and explore the future development of synaesthesia integration, this project builds a set of indoor human activity perception methods using SRS signals based on the RAN of OpenAirInterface5G, which can achieve real-time human activity perception within 10ms, reaching 99 in six tasks.More than 9% recognition accuracy.
-Used for motion recognition demonstration in action gamesvideos\ActionControl-SFV.mp4. In this demo, the characters will perform punches, left and right movements, squats, kicks, and exquisite combat power actions to correspond to the actions of the characters in the fighting game.
There is also a demo videos\2077.mp4 that uses other custom actions for training and perception in driving games, where Cyberpunk 2077 is selected for display.：
The perception algorithm is effective. We applied the same algorithm to the base station of H3C, and the corresponding motion recognition can also be achieved. See videos\H3C.mp4 for details. The white square box of the mobile cabinet is the RRU.
Please note: there was no open source plan for the work of this article at the beginning, so there are many irregularities and improprieties in the commit, please forgive me. The author no longer conducts research in the field of communication, so he decided to make these works public to help follow-up researchers accelerate related research and promote the development of communication, but the corresponding author will not maintain these codes. If you need to learn more about the business cooperation or visit of this project, you can contact Sony China Research Institute (Beijing) to negotiate. I would like to thank all the leaders and colleagues of Sony China Research Institute for their support and cooperation.
This project is divided into three stages：
E-Mail: cybersh1t@126.com
Github: @Dafeigy
The related work of this project won the Sony 2023 IISC Ownership Award, and accepted the Interview of China International Television, a subsidiary of CCTV ：
The project is cooperating with large domestic base station equipment manufacturers. The perception algorithm has been successfully integrated with commercial base station products, and is further expanding, exploring the scope and accuracy of perception, and considering the development of a new generation of base station products.
The work of this project was carried out by me as the leading developer in the horizontal project of participating in the Sony China Research Institute. The data collection and perceptual algorithm testing of large-scale conference room scenes and classroom scenes were carried out at the Sony China Research Institute. Except for the author himself, the remaining testers are all employees of Sony China Research Institute. Under the condition that it does not violate the relevant cooperative interests and the attribution of intellectual property rights, only the following content is open source：
-Code implementation of FreqBlock/FreqNet. Refer to SRS-Front.
-OpenAirInterface5G+Free5GC+ commercial UE connection method. Refer to my blog.
-OAI code for secondary development. Please pay attention to the relevant terms of use of the OAI code.
-Implementation of channel estimation SIMD acceleration. Please check Warehouse link In the OAI-test branch. You can make corresponding modifications according to the number of subcarriers you use and the antenna configuration when the UE is connected.
-MINI-KM that implements the control of “somatosensory control” games in the Windows platform.
-Visualization platform SRS-Front.
-Perceive the relevant data of the “author himself” in the data set. Due to the huge amount of data, if you need to use the author's own data, you can contact the author: cybersh1t@126.com. The open source link only provides a small amount of data for readers to view the data set style；
-Based on OAI collection of CSI data, Data collection and processing script and other miscellaneous content.
-Build an experimental platform for Free5GC (or other core networks) + OAI RAN + commercial UE. Refer to my blog Article: OAI environment configuration, Free5GC, Free5GC+OAIRAN deployment.
-Modify the relevant code of OAI RAN.
-Collect data in the experimental site and train the model, and use data processing scripts to process the data.
-Load the perception model in SRS-Front and reason accordingly, or use MINI-KM to “somatosensory control” the game.