How You Distinguish People by Voice

Pu Fanyi*, Jin Qingyang*, Soo Ying Xi*, Shan Yi*, Zhang Xintong*
Nanyang Technological University, Singapore
* means equal contributions
Report Code HF Logo HuggingFace Dataset

People can generally distinguish the characteristics of a speaker by their voice. This project investigates how people identify others through specific features of certain vocal signals. We have released a new dataset, DiffVoice, and explored the differences in voices emitted by different individuals using various statistical analysis methods.


This project is for the course MH3511 Data Analysis with Computer, AY2023/24 Semester 2.

The DiffVoice Dataset

The DiffVoice dataset is a collection of voice recordings from different individuals. The dataset contains various features extracted from the voice signals, such as pitch, formants, and MFCCs. The dataset is designed to help researchers explore the differences in voices emitted by different individuals and develop models to distinguish between them.

The dataset is available on HuggingFace, while a CSV version can be downloaded here.

DiffVoice Dataset