Voice Device Individual Source Identification Dataset

Several phone models are represented by at least two individual devices.The recordings in the dataset cover diverse scenarios such as silence, daily conversation, multi-person dialogue, video recording, and background noise. The recording environments include multiple real-world settings including classrooms, dormitories, high-speed trains, laboratories, and track fields.Each individual mobile device contains no fewer than 300 recording clips, each with a duration of 3 seconds and a speech sampling rate of 16 kHz.The 121 mobile device individuals in the dataset were randomly divided into two parts: 106 of them were used as the training set, and the remaining 15 as the test set, ensuring no overlap in mobile phone models between the training and test sets.For the training set, each audio recording was clipped to a duration of 3 seconds, read at a sampling rate of 16000 Hz, and used for feature extraction.

Each mobile phone model contained no fewer than 300 speech segments.For the test set, random pairings were generated from the 15 models. Recording pairs originating from the same mobile device individual were labeled as 1, while those from different individuals were labeled as 0, thus constructing the test set .