Summary

​The Vocal Distance Dataset consists of individuals who were recorded speaking in Korean, which amounted to a total of 573 hours. Due to the placement of recording devices in three different distances from the individuals, there is 191 hours of audio for each distance: 0.4m, 2m, and 4m. Since sound varies widely with distance, this dataset will allow engineers to build more accurate models that can identify sounds. Instead of relying on a dataset recorded from one distance, it will be more reliable to use a dataset that contains multiple distances depending on how far away the speaker is from the microphone. To record the data, there were three microphones placed at a distance of (0.4m, 2m, and 4m) away from the speaker. To maintain the consistency of the data, the speaker was always standing and the average height of each microphone was 1.5m tall.