毫米波室内步态数据集

步态识别，即从人的行走姿势识别人，已在安全检查、健康监测和新型人机交互中得到了广泛的应用。基于毫米波（mmWave）的步态识别代表了最近的进步。与传统的基于摄像头的解决方案相比，基于毫米波的步态识别具有独特的优势，在非视线场景下仍然有效，例如在黑暗、弱光或阻塞条件下。此外，他们还能够在保护隐私的同时进行身份识别。目前，由于没有可用的公共数据集，mmWave步态识别的工作很少。在本文中，我们构建了第一个同类的毫米波步态数据集，其中我们收集了95名志愿者在两种不同场景下从两个毫米波雷达中“看到”的步态，这两种场景总共持续约30小时。

Gait recognition, which identifies individuals based on their walking postures, has been widely applied in security inspection, health monitoring and novel human-computer interaction scenarios. Millimeter-wave (mmWave)-based gait recognition represents recent advancements in this field. Compared with traditional camera-based solutions, mmWave-based gait recognition has unique advantages: it remains effective in non-line-of-sight (NLOS) scenarios such as darkness, low-light environments or obstructed conditions. Furthermore, it enables identity recognition while preserving user privacy. Currently, research on mmWave-based gait recognition is limited due to the lack of publicly available datasets. In this paper, we construct the first-of-its-kind mmWave gait dataset, where we collected gait data from 95 volunteers captured by two mmWave radars across two different scenarios, with a total duration of approximately 30 hours.

搜集汇总

背景与挑战

背景概述

该数据集是首个公开的毫米波室内步态数据集，专注于步态识别应用，具有在非视线场景（如黑暗或阻塞条件下）有效且保护隐私的优势。数据集包含95名志愿者在两种场景下的步态数据，总时长约30小时，数据量4.39GB，适用于安全检查、健康监测和人机交互等领域的研究。