融合呼吸信号和运动信号的人体行为识别方法

Human behavior recognition method based on fusion of respiratory signal and motion signal

人体行为识别技术在许多领域都有广泛的运用,但是目前基于仍然存在传感器类型单一,适用场景有限,捕捉人体活动时特征不足的问题。本文提出了一种可穿戴的多模块传感器的人体行为识别方案,使用惯性传感器和拉伸应变传感器。拉伸应变传感器穿戴于胸前,能够有效的捕捉到不同行为下的人体呼吸信号。同时,本文使用了多尺度的一维卷积神经网络,能够有效的提取传感器数据的特征。实验结果表明,在本文提出的7种活动中,多尺度一维卷积神经网络达到了96.89%的识别率,说明了本文所提出的方法能够很好地用于人体行为识别任务之中。

Human behavior recognition technology has been widely used in many fields,but there are still some problems such as single sensor type,limited applicable scenes and insufficient fea⁃tures when capturing human activities.This paper proposes a wearable,multi module sensor human behavior recognition scheme,using inertial sensor and tensile strain sensor.The tensile strain sen⁃sor is worn on the chest,which can effectively capture the human respiratory signals under different behaviors.At the same time,this paper uses a multi-scale one-dimensional convolutional neural network,which can effectively extract the characteristics of sensor data.The experimental results show that the recognition rate of multi-scale one-dimensional convolutional neural network reaches 96.89%in the seven activities proposed in this paper,which shows that the method proposed in this paper can be well used in the task of human behavior recognition.