BanglaMood: A Rich Audio Dataset for Eight Emotion Recognition in Bangla Speech

The BanglaMOOD dataset comprises 3,220 audio files in WAV format, each capturing vocal expressions of human emotion in Bangla. The dataset features eight emotion categories: Angry, Happy, Sad, Neutral, Surprised, Fear, Disappointment, and Excitement. Each audio sample ranges from 1 to 5 seconds in duration, with all files standardized to 5 seconds for consistency in feature extraction. The audio was collected from a wide variety of real-life contexts, including short stories, vlogs, and sports highlights sourced from YouTube, ensuring a broad range of expressive and natural vocalizations. Both male and female voices are included to provide gender diversity. In terms of distribution, the Angry emotion has the highest number of samples, followed by Neutral, while Excitement has the lowest. Specifically, there are 300 Happy samples (179 male and 121 female), 337 Sad samples (148 male and 189 female), 611 Angry samples (402 male and 209 female), 610 Neutral samples (477 male and 133 female), 303 Surprise samples (206 male and 97 female), 449 Fear samples (348 male and 101 female), 343 Disappointment samples (228 male and 115 female), and 267 Excitement samples (202 male and 65 female). In most categories, male voices are more prevalent, except for the Sad emotion, where female samples are more numerous. Angry emotion has the highest number of samples for females. Overall, the BanglaMOOD dataset offers a diverse collection of emotional speech samples in Bangla, making it a valuable resource for research and development in speech emotion recognition and related fields.