PoseKernel Dataset

PoseKernel数据集组成了针对六个不同环境的各种3D人体姿势的声学冲动响应数据。我们在环境中放置了四对共同定位的扬声器和麦克风，以及一个Kinect相机来捕获groundtruth 3D姿势。我们释放每对扬声器和麦克风的声学冲动响应，以及麦克风和扬声器的位置，以及groundtruth 3D姿势。所有与位置相关的信息都在相机的框架内。The PoseKernel dataset consists of acoustic impulse response data of various 3D human poses across six distinct environments. We deployed four pairs of co-located speakers and microphones, as well as a Kinect camera, within these environments to capture ground-truth 3D human poses. We release the acoustic impulse responses for each speaker-microphone pair, along with the positions of the microphones and speakers, and the ground-truth 3D poses. All position-related information is expressed within the coordinate frame of the Kinect camera.

搜集汇总

背景与挑战

背景概述

PoseKernel数据集是一个包含六个环境下的3D人体姿势声学冲动响应数据集合，通过扬声器-麦克风对和Kinect相机捕获groundtruth姿势，并提供相关位置和响应信息。该数据集由多所大学和三星AI中心于2022年发布，用于声学与姿势分析研究。