DISC: a Dataset for Integrated Sensing and Communication in mmWave Systems

TL;DR

DISC is a comprehensive mmWave dataset capturing channel responses for integrated sensing and communication, enabling advancements in human activity recognition and RF sensing with real Wi-Fi traffic patterns.

Contribution

First dataset of its kind using a software-defined radio testbed with real Wi-Fi traffic for integrated sensing and communication in mmWave systems.

Findings

Dataset enables benchmarking of sensing algorithms

Open-source baseline algorithms demonstrate utility

Data supports multiple human sensing applications

Abstract

This paper presents DISC, a dataset of millimeter-wave channel impulse response measurements for integrated human activity sensing and communication. This is the first dataset collected with a software-defined radio testbed that transmits 60 GHz IEEE 802-11ay-compliant packets and estimates the channel response including scattered signals off the moving body parts of subjects moving in an indoor environment. The provided data consists of three parts, for more than 2 hours of channel measurements with high temporal resolution (0.27 ms inter-packet time). DISC contains the contribution of 7 subjects performing 5 different activities, and includes data collected from two distinct environments. Unlike available radar-based millimeter-wave sensing datasets, our measurements are collected using uniform packet transmission times and sparse traffic patterns from real Wi-Fi deployments. We develop, train, and release open-source baseline algorithms based on DISC to perform human sensing tasks. Our results demonstrate that DISC can serve as a multi-purpose benchmarking tool for machine learning-based human activity recognition, radio frequency gait analysis, and sparse sensing algorithms for next-generation integrated sensing and communications.