Scholarlink

Text stands to be one of the prominent aspects of Communication in any Technical Phenomenon. But, with Text, understanding the Emotion of a Person, which plays a vital role in decision-making for actuating a Task, isn’t possible. With the Escalated Indulgence of Speech into Diverse Technological Systems, it has been necessary to formulate a modus operandi which would precisely determine the Emotion of a Person through its Speech. Different approaches have been proposed for recognizing the Emotion of a Person through Speech, but the Limitation of Modular Outlook over Speech Data seems to be Visible. In this paper, we consummated Convolutional Neural Networks and tried to understand the Impact of the Convolution Layers, i.e., 03, 04, 05, and 06—Layered CNN, over the Speech Data and thereby formulate a Potent and Efficient Solution for Speech Emotion Recognition. We utilized the well-known Open-Source Datasets such as CREMA, RAVDESS, SAVEE, and TESS, consisting of Speech Audio for various Human Emotions, and combined them into one whole Dataset. The outcome we got was phenomenal as the Accuracy surpassed the already existing methodologies, and also it gave us an Impactful Elucidation of its Real-time Applicability for varied use-cases related to Speech.

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

Leveraging Differently Layered Convolutional Neural Network for Speech Emotion Recognition

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Prathik Kumar Gaddam,
  • Anas Iqubal,
  • Muqtada Hasan,
  • Arun Prakash Agrawal

摘要

Text stands to be one of the prominent aspects of Communication in any Technical Phenomenon. But, with Text, understanding the Emotion of a Person, which plays a vital role in decision-making for actuating a Task, isn’t possible. With the Escalated Indulgence of Speech into Diverse Technological Systems, it has been necessary to formulate a modus operandi which would precisely determine the Emotion of a Person through its Speech. Different approaches have been proposed for recognizing the Emotion of a Person through Speech, but the Limitation of Modular Outlook over Speech Data seems to be Visible. In this paper, we consummated Convolutional Neural Networks and tried to understand the Impact of the Convolution Layers, i.e., 03, 04, 05, and 06—Layered CNN, over the Speech Data and thereby formulate a Potent and Efficient Solution for Speech Emotion Recognition. We utilized the well-known Open-Source Datasets such as CREMA, RAVDESS, SAVEE, and TESS, consisting of Speech Audio for various Human Emotions, and combined them into one whole Dataset. The outcome we got was phenomenal as the Accuracy surpassed the already existing methodologies, and also it gave us an Impactful Elucidation of its Real-time Applicability for varied use-cases related to Speech.