Scholarlink

<p>In this paper, a biologically inspired algorithm is proposed for emotional content recognition of speech. Our method can recognize seven emotional states, namely, Anger, Boredom, Disgust, Fear, Happiness, Sadness and Neutral. The inspiration for our algorithm is based on the mammalian brains specifically the human brain and is commonly known as the Brain Emotional Learning (BEL). The BEL model consists of four parts, Thalamus, Sensory Cortex, Orbitofrontal Cortex (OFC) and Amygdala. In our method, we have used Softmax function for the implementation of the OFC and Amygdala. In this study, two feature extraction methods are used and studied independently for the representation of the features that carry the emotional content. These techniques are Short-time Fourier Transform (STFT) and Mel-frequency Cepstral Coefficients (MFCC). Our experimental results show that the proposed algorithm can recognize seven emotional states with a good rate of accuracy. In our testing, the STFT method shows a better performance compared to the MFCC coefficients method.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Speech emotional content recognition based on the brain emotional learning model

  • 期刊论文

  • 发表日期：

    2026年2月2日

查看全文
  • Mahsa Ravanbakhsh,
  • Saeed Setayeshi,
  • Mohammad Ravanbakhsh

摘要

In this paper, a biologically inspired algorithm is proposed for emotional content recognition of speech. Our method can recognize seven emotional states, namely, Anger, Boredom, Disgust, Fear, Happiness, Sadness and Neutral. The inspiration for our algorithm is based on the mammalian brains specifically the human brain and is commonly known as the Brain Emotional Learning (BEL). The BEL model consists of four parts, Thalamus, Sensory Cortex, Orbitofrontal Cortex (OFC) and Amygdala. In our method, we have used Softmax function for the implementation of the OFC and Amygdala. In this study, two feature extraction methods are used and studied independently for the representation of the features that carry the emotional content. These techniques are Short-time Fourier Transform (STFT) and Mel-frequency Cepstral Coefficients (MFCC). Our experimental results show that the proposed algorithm can recognize seven emotional states with a good rate of accuracy. In our testing, the STFT method shows a better performance compared to the MFCC coefficients method.