机器人自身噪声环境下的自动语音识别
发布时间：2019-06-05 08:34
【摘要】：当机器人移动身体任何部位时,都会不可避免地产生自身噪声。这些自身噪声由身体关节或其他硬件设备如风扇等引起。由于自身噪声距离机器人麦克风较近,较目标声源更容易被获取。该文根据机器人自身噪声种类,提出了一种将谱减法、关节噪声模板减法、基于标注区域的倒谱均值减法以及多条件训练相结合的方法,从而估计和抑制自身噪声。一系列实验证明了所提出的方法可以有效地减少自身噪声影响,提高语音识别的鲁棒性。
[Abstract]:When the robot moves any part of the body, it will inevitably produce its own noise. These self-noise is caused by body joints or other hardware devices such as fans. Because its own noise is close to the robot microphone, it is easier to obtain the target sound source than the target sound source. According to the type of robot noise, this paper proposes a method which combines spectral subtraction, joint noise template subtraction, cepstrum mean subtraction based on marked region and multi-condition training to estimate and suppress its own noise. A series of experiments have proved that the proposed method can effectively reduce the influence of noise and improve the robustness of speech recognition.
【作者单位】： 天津大学计算机科学与技术学院;天津大学软件学院;
【基金】：国家自然科学基金资助项目(61471259;61304250;61573254)
【分类号】：TP242;TN912.34
[Abstract]:When the robot moves any part of the body, it will inevitably produce its own noise. These self-noise is caused by body joints or other hardware devices such as fans. Because its own noise is close to the robot microphone, it is easier to obtain the target sound source than the target sound source. According to the type of robot noise, this paper proposes a method which combines spectral subtraction, joint noise template subtraction, cepstrum mean subtraction based on marked region and multi-condition training to estimate and suppress its own noise. A series of experiments have proved that the proposed method can effectively reduce the influence of noise and improve the robustness of speech recognition.
【作者单位】： 天津大学计算机科学与技术学院;天津大学软件学院;
【基金】：国家自然科学基金资助项目(61471259;61304250;61573254)
【分类号】：TP242;TN912.34
【相似文献】
相关期刊论文 前10条
1 陈林;奚如如;王兴松;;套索驱动细长机器人的初步设计与试验[J];机电工程;2011年03期
2 谢芝馨;;苏联机器人技术述评[J];机械与电子;1989年04期
3 罗飞，余达太;主动式控制──机器人抑振控制的有效方式[J];机器人;1995年04期
4 卢桂章;当前高技术发展的前沿——机器人技术[J];天津科技;1995年01期
5 苏陆;日本机器人技术与产品[J];全球科技经济w，
本文编号：2493389
本文链接：https://www.wllwen.com/kejilunwen/zidonghuakongzhilunwen/2493389.html
