sig_competingtalkers(signame) loads one of several test signals consisting of competing talkers. All the talkers are taken from the TIMIT speech corpus https://doi.org/10.35111/17gk-bn40 and filtered by HRTFs recorded in the Oldenburg lab with a Bruel and Kjaer Type 4128C head and torso simulator (Kayser et al., 2009). An exception is 'one_speaker_reverb', see description below.
The signals have 2 channels and are provided with a sampling rate of 16 kHz.
[sig,fs]=sig_competingtalkers(signame) additionally returns the sampling frequency fs.
The parameter signame can be:
|'one_of_three'||Talker spatialized at the azimuth angle of 30 degrees.|
|'two_of_three'||Talker spatialized at the azimuth angle of 0 degrees.|
|'three_of_three'||Talker spatialized at the azimuth angle of -30 degrees.|
|'one_speaker_reverb'||Talker spatialized at horizontal position of 45 degrees by applying a binaural set of room impulse responses (BRIR) of an office room from a database recorded with hearing-aid microphones behind the ear.|
|'two_speakers'||Two simultanous talkers spatialized at the azimuth angles of -30 and 30 degrees.|
|'five_speakers'||Five simultanous talkers spatialized at the azimuth angles of -80, -30, 0, 30, and 80 degrees.|
|'bnoise'||Speech-shaped binaural noise.|
The following plot shows an estimate of the power spectral density of the first channels of the speech shaped noise:
This code produces the following output:
Model: sig_competingtalkers. Downloading auxiliary data: bnoise.wav Searching in current AMT version... Searching in previous AMT versions... Trying version amt-1.0.0 Found data in version: amt-1.0.0