Skip to main content

Table 1 Characteristics and statistics of the DIRHA-sim and DIRHA-real corpora, used in the experiments of this paper

From: Room-localized speech activity detection in multi-microphone smart homes

Data Databases
characteristics DIRHA-sim DIRHA-real
Speech source Loudspeaker Human
1-min-long sequences (#) 150 60
Total speech (min) 47 19
Overlapped speech (min) 22 0
Non-speech events (#) 72 Untranscribed
Background noises (#) 10 Untranscribed
Subjects (#) 20 5
Average SNR (dB) 13 15