人头
稳健性(进化)
计算机科学
主管(地质)
人耳
旋转(数学)
计算机视觉
人工智能
声学
物理
地质学
生物化学
基因
地貌学
吸收(声学)
化学
作者
Y. J. Liu,Haowen Li,Haishan Zou,Zhibin Lin,Jing Lü
摘要
Due to the limited size of the quiet zone created by active headrests (AHR) near the human ear, noise reduction (NR) at the human ear decreases dramatically when the head moves. Combined with a head tracking system can improve the NR performance when the head moves, but most such studies currently only consider head translation. To improve the robustness when the head translates or rotates, an ear-positioning (EP) system based on a depth camera and human pose estimation model is presented in this paper and integrated with AHR. A post processing method is proposed to address extreme scenarios like an ear being occluded. Experimental results show that the EP system can effectively track the movement of ears. The performance of AHR combined with the system is more robust, achieving the lowest 11.7/12.2 dBA (left ear/right ear) NR for white noise in range of 80–2000 Hz when head translates in a 5 × 5 × 2 grid at a 2.5 cm interval and 11.4/13.6 dBA for head rotation within the range of 60° compared to −5.7/−6.9 and −3.7/2.5 dBA without the system.
科研通智能强力驱动
Strongly Powered by AbleSci AI