.png){kind=icon}
返回
A Smartphone Pose Auto-calibration Method using Hash-based DOA Estimation
.png){kind=icon}
PRE
AI
DOI：10.1109/SII55687.2023.10039085.png)
.png){kind=icon}
原文求助.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
This paper presents a method to utilize multiple off-the-shelf smartphones to localize speakers. For DOA (direction of arrival) estimation on every single smartphone, we proposed an O(1) complexity hash table-based modified phase transform (PHAT) estimation method without scanning all possible directions to achieve lower CPU usage and longer battery life. Additionally, to increase DOA estimation accuracy, we measured two types of smartphone impulse responses and made them publicly available. In the auto-calibration process, each smartphone detects a pure tone emitted from another smartphone's speaker. Assuming that all smartphones are on the same desktop surface, each smartphone's 2D position and rotation are estimated using these detected DOAs and the speaker position relative to their central point. A bundle adjustment-like optimization method is employed to reduce the re-projection error in this process. After auto-calibration, we can easily integrate the DOAs found by each smartphone and estimate the speaker's position using simple triangulation. The experimental results show that the proposed hash table-based DOA estimation method and 2D version bundle adjustment can perform auto-calibration precisely.
Keyword:
PARAMETERS
MULTIPLE
ESPRIT
期刊
I
IF:
0
论文数:
5
被引数:
0
机构
暂无机构信息
引用论文
.png){kind=icon}
分享收藏
查看更多内容