Bengali Vocal Spectrum (BVS): A Bengali Voice Dataset for Psychological Stability Analysis

The Bengali Vocal Spectrum (BVS) dataset is a valuable resource for psychological stability analysis, featuring 85 audio recordings in the Bengali language, totaling approximately 4.65 hours. The dataset is divided into two groups: 43 recordings (2.63 hours) from psychologically unstable patients and 42 recordings (2.02 hours) from psychologically stable individuals. These recordings were collected at Pabna Mental Hospital, Pabna, and National Institute of Mental Health & Hospital, Dhaka. The dataset was compiled with a firm commitment to ethical standards, including obtaining all necessary permissions and ensuring the confidentiality and anonymity of all participants. The data was meticulously preprocessed using Audacity software to normalize the volume, remove background noise, and ensure clarity, with each file saved in .wav format at a 48000 Hz sample rate. Spectrogram images are in a separate folder and segmented into 2-second clips from the audio data for signal processing techniques, enhancing the utility of the dataset for machine learning applications. Researchers can utilize this dataset to develop or refine algorithms that assess psychological stability based on vocal characteristics. The data's format and quality also make it suitable for advanced signal processing tasks, which could lead to new insights into the acoustic markers associated with psychological conditions.