论文部分内容阅读
为了有效识别自发、非典型及未分割语音的情感以建立更自然的人机交互界面,提出了一种新的结合情感数据场和蚁群策略的语音情感识别算法.用情感数据场中势函数建立基于块的声学特征向量之间的内在联系.为识别自发语音情感,用人工蚁群模拟基于块的声学特征向量,然后用典型的蚁群策略研究每个人工蚂蚁在情感数据场的运动轨迹,并把该蚂蚁的运动轨迹作为对应的声学特征向量的情感标签.利用2012年连续音视频情感挑战赛中的语音数据对所提算法进行测试.实验结果表明:该算法较已有算法能更好地对基于块的语音情感进行识别.
In order to identify spontaneous, atypical and undivided emotions and establish a more natural human-computer interaction interface, a novel speech emotion recognition algorithm based on emotion data field and ant colony strategy is proposed. To establish the intrinsic relationship between block-based acoustic feature vectors.In order to identify spontaneous speech emotion, artificial ant colony is used to simulate block-based acoustic eigenvectors, and then the ant colony strategy is used to study the motion trajectory of each artificial ant in the emotion data field, And use the motion trajectory of the ant as the emotion label of the corresponding acoustic eigenvector.The proposed algorithm is tested by using the speech data in the continuous audio and video challenge competition in 2012. The experimental results show that the algorithm is better than the existing algorithm Based on the block to identify the emotional voice.