IndoWaveSentiment: Indonesian Audio Dataset for Emotion Classification

IndoWaveSentiment is an audio dataset designed for classifying emotional expressions in Indonesian speech. The dataset was created using recordings from 10 actors, equally split between men and women. Recordings were made using a mono channel cardioid vocal microphone positioned no more than 10 cm from the speakers, connected to a laptop or computer. The audio was captured at a sample rate of 16 kHz with 32-bit depth. Each actor repeated the same sentence three times across five emotional categories: neutral, happy, surprised, disgusted, and disappointed. The dataset comprises 300 .wav files in total. This data was utilized to develop deep learning models as an initial study on audio classification in multimodal sentiment analysis. IndoWaveSentiment is also suitable for various signal processing applications, including voice emotion classification, and supports the development of sentiment analysis.

IndoWaveSentiment是一款专为印尼语语音情感表达分类任务设计的音频数据集（audio dataset）。该数据集由10名演员的录音构建而成，男女演员各占半数。录制时采用单通道心形人声麦克风（mono channel cardioid vocal microphone），麦克风与发声者的间距不超过10厘米，设备连接至笔记本电脑或台式计算机。音频采集参数为采样率16 kHz，位深度32位。每位演员需针对中性、开心、惊讶、厌恶、失望这五类情感类别，各重复录制同一句话三次。该数据集总计包含300个.wav格式音频文件。本数据集曾被用于开发深度学习模型，作为多模态情感分析（multimodal sentiment analysis）领域音频分类方向的一项初步研究。此外，IndoWaveSentiment同样适用于包括语音情感分类在内的各类信号处理应用，可为情感分析相关研发工作提供支撑。