Scholarlink

<p>The Whispered Speech Emotion Recognition (SER) presents special challenges because it does not involve the vibration of vocal folds and less acoustic information, which makes it more difficult to identify the nuances of emotions and gender-related peculiarities than in normal speech. The research presents a new system, HybridOpti-GenderNet, of concurrent whispered speech emotion and gender recognition. It is a framework of six stages are data collection, pre-processing, feature extraction, feature selection, gender detection, and emotion recognition (ER). A Red Butterfly Optimization Algorithm (RBOA), which uses Butterfly Optimization Algorithm (BOA) and Red Kite Optimization Algorithm (ROA) by adjusting the weight to achieve optimal feature selection, is proposed to achieve a better balance between exploration and exploitation. Gender identification is carried out with the proposed HybridOpti-GenderNet, a new hybridization of Bi-LSTM and DCNN, and ER is accomplished with the help of the Attention-Enhanced Hybrid Belief Network (AHBN) that incorporates Deep Belief Network (DBN), Feedforward Neural Network (FNN) &amp; a tailor-made attention mechanism to detect delicate emotional signals. The innovativeness of the given work is that the gender-informed recognition is combined with a high-level optimization and a hybrid deep learning pipeline that enhances the robustness of the analysis of whispered speech significantly. The system was tested in Python and tested on the GeWEC and EMO-DB data, showing superior performance in cases of privacy preservation, noisy conditions, and real-life human-computer interaction.</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

Whispered speech emotion recognition with gender detection using hybridopti-gendernet

  • 期刊论文

  • 发表日期：

    2026年3月30日

查看全文
  • Aniruddha Mohanty,
  • Ravindranath Chowdary Cherukuri

摘要

The Whispered Speech Emotion Recognition (SER) presents special challenges because it does not involve the vibration of vocal folds and less acoustic information, which makes it more difficult to identify the nuances of emotions and gender-related peculiarities than in normal speech. The research presents a new system, HybridOpti-GenderNet, of concurrent whispered speech emotion and gender recognition. It is a framework of six stages are data collection, pre-processing, feature extraction, feature selection, gender detection, and emotion recognition (ER). A Red Butterfly Optimization Algorithm (RBOA), which uses Butterfly Optimization Algorithm (BOA) and Red Kite Optimization Algorithm (ROA) by adjusting the weight to achieve optimal feature selection, is proposed to achieve a better balance between exploration and exploitation. Gender identification is carried out with the proposed HybridOpti-GenderNet, a new hybridization of Bi-LSTM and DCNN, and ER is accomplished with the help of the Attention-Enhanced Hybrid Belief Network (AHBN) that incorporates Deep Belief Network (DBN), Feedforward Neural Network (FNN) & a tailor-made attention mechanism to detect delicate emotional signals. The innovativeness of the given work is that the gender-informed recognition is combined with a high-level optimization and a hybrid deep learning pipeline that enhances the robustness of the analysis of whispered speech significantly. The system was tested in Python and tested on the GeWEC and EMO-DB data, showing superior performance in cases of privacy preservation, noisy conditions, and real-life human-computer interaction.