Emotion recognition from speech signal
Emotion recognition from speech signal
复制标题
从语音信号中进行情感识别
DOI：
--
复制
发表时间：
2017
期刊：
影响因子：
--
通讯作者：
V. K. Mittal
中科院分区：
文献类型：
--
作者：
Esther Ramdinmawii;Abhijit Mohanta;V. K. Mittal
情感识别是近年来发展迅速的一个研究领域。与人类不同，机器缺乏感知和展示情感的能力。但通过自动情绪识别可以改善人机交互，从而减少人工干预的需要。本文从情感语音信号出发，分析了四种基本情绪(愤怒、快乐、恐惧和中性)。使用信号处理方法从这些信号中获取产品特征。信号源特征为瞬时基频(F0)，系统特征为共振峰和主频，过零率(ZCR)，综合特征信号能量用于分析。F0采用零频滤波，共振峰和主频采用LP谱。使用200个样本的矩形窗口，在浊音和清音区域中获得短时信号能量(STE)和ZCR。两个数据库，德语和Telugu情感数据库被用来交叉验证结果。在高唤醒情绪(愤怒和快乐)和中性情绪之间观察到明显的差异。结果表明，愤怒和快乐情绪之间存在重叠。但在快乐/愤怒和恐惧，以及快乐和愤怒情绪之间的特征上观察到了明显的不同，否则这是一个具有挑战性的问题。所获得的见解可能会在一系列应用中有所帮助。
Emotion recognition is a rapidly growing research domain in recent years. Unlike humans, machines lack the abilities to perceive and show emotions. But human-computer interaction can be improved by automated emotions recognition, thereby reducing the need of human intervention. In this paper, four basic emotions (Anger, Happy, Fear and Neutral) are analyzed from emotional speech signals. Signal processing methods are used for obtaining the production features from these signals. Source feature the instantaneous fundamental frequency (F0), system features the formants and dominant frequencies, zero-crossing rate (ZCR), and the combined features signal energy are used for the analyses. F0 is obtained using zero-frequency filtering (ZFF), and formants and dominant frequencies using LP spectrum. Short-time signal energy (STE) and ZCR are obtained in the voiced and unvoiced regions using a rectangular window of 200 samples. Two databases, German and Telugu Emotion Databases are used to cross-validate the results. Distinct differences are observed between high-arousal emotions (Anger and Happy) and Neutral emotion. Results indicate overlap between Anger and Happy emotions. But distinct differences are observed in the features for Happy/Anger and Fear, and between Happy and Anger emotions which is otherwise a challenging problem. The insights gained may be helpful in range of applications.