In this article, the new cognitive model convolutional brain emotional learning (CBEL) is introduced to recognize emotional speech on the Berlin dataset. This model is an improved model of brain emotional learning (BEL), which is inspired by the limbic function of the brain. The reason for choosing this model is that the limbic system of the brain is responsible for emotion, so it can help to recognize the emotion of speech. In the proposed method the input signals will be sent to the convolutional neural network (CNN) for the feature extraction. Then, changes have been made in the CNN pooling layer to increase the processing speed. Finally, the output of the pooling layers was sent to the Multi-Layer Perceptron (MLP) networks in the amygdala and orbitofrontal, from the CBEL model, to display the recognition rate on each basic emotional state. The results of the experiments show that the accuracy of emotional speech recognition of the proposed model is 97.39% and it has performed better than other methods introduced in the article.

在本文中，引入了新的认知模型卷积脑情感学习（CBEL）来识别柏林数据集上的情感语音。该模型是大脑情感学习（BEL）的改进模型，其灵感来自于大脑的边缘系统功能。选择这个模型的原因是大脑的边缘系统负责情感，因此可以帮助识别言语的情感。在所提出的方法中，输入信号将被发送到卷积神经网络（CNN）进行特征提取。然后，对 CNN 池化层进行了更改以提高处理速度。最后，池化层的输出从 CBEL 模型发送到杏仁核和眶额的多层感知器 (MLP) 网络，以显示每个基本情绪状态的识别率。实验结果表明，该模型的情感语音识别准确率为97.39%，表现优于文中介绍的其他方法。