dnzblgn/Voice_Sentiment_Analysis

Open Source JL-corpus was used as the primary dataset for this dataset. The corpus includes 300 short audio files, 150 of which were recorded by two different females and 150 of which were recorded by two different males in New Zealand English. Each audio file is labeled with one of five emotions: angry, happy, neutral, excited, and sad. The transcription, sex, F0, maximum pitch, minimum pitch, F1, F2, F3, intensity, maximum intensity, minimum intensity, duration, and emotion label were extracted for each audio file using Praat.

原始信息汇总

数据集概述

数据来源

• 本数据集主要基于开源的JL-corpus。

数据内容

• 包含300个短音频文件。
• 其中150个音频由两位不同女性录制，另外150个音频由两位不同男性录制，均使用新西兰英语。

音频特征

• 每个音频文件均标注了以下特征：
  • 转录文本
  • 说话人性别
  • 基频（F0）
  • 最大音高
  • 最小音高
  • 频谱特征（F1, F2, F3）
  • 强度
  • 最大强度
  • 最小强度
  • 持续时间
  • 情感标签（愤怒、快乐、中性、兴奋、悲伤）

数据处理

• 使用Praat软件提取了上述所有音频特征。