# Digits-in-Noise Perceptual Dataset

Experimental Data: Interference of mid-level sound statistics predicts human speech recognition in natural noise

DOI: https://doi.org/10.1101/2024.02.13.579526


Experiment 1: Original (OR), Phase Randomized (PR), Spectrum Equalized (SE) Data
- 825 Foreground Sounds
- 825 Background Sounds
- BehavioralDataExp1.mat
    - BackSoundNum: (1-11) as shown in paper (1x825 trials)
    - BackCondition: 'OR' 'PR' 'SE' (1x825 trials)
    - SNR: Signal-to-Noise-Ratio Value Between Fore and Back  (1x825 trials)
    - Resp: Behavioral Response (18 participants x 825 trials) - 0 (incorrect response), 1 (correct response)

Experiment 2: Gradually Added Texture Statistics in Babble8 and Jackhammer (Behavioral Response Only)
- BehavioralDataExp2_Babble8.mat
    - BackSoundNum: (1-11) as shown in paper (1x500 trials)
    - SNR: Signal-to-Noise-Ratio Value Between Fore and Back (1x500 trials)
    - TextureStat: {SPEC,MAR,MOD,CORR,ORIG} representing gradually added texture statistics (McDermott 2011)
    - Resp: Behavioral Response (10 participants x 500 trials) - 0 (incorrect response), 1 (correct response)
    
- BehavioralDataExp2_Jackhammer.mat
    - BackSoundNum: (1-11) as shown in paper (1x500 trials)
    - SNR: Signal-to-Noise-Ratio Value Between Fore and Back (1x500 trials)
    - TextureStat: {SPEC,MAR,MOD,CORR,ORIG} representing gradually added texture statistics (McDermott 2011)
    - Resp: Behavioral Response (6 participants x 500 trials) - 0 (incorrect response), 1 (correct response)

Experiment 3: Varying SNR Across OR Conditions
- V1\_SNR\_-15\_-3
    - 1375 Foresounds
    - 1375 Backsounds
    - BehavioralDataExp3_v1.mat
        - BackSoundNum: (1-11) as shown in paper (1x1375 trials)
        - SNR: Signal-to-Noise-Ratio Value Between Fore and Back (1x1375 trials)
        - Resp: Behavioral Response (5 participants x 1375 trials) - 0 (incorrect response), 1 (correct response)

- V1\_SNR\_-18\_0
    - 1925 Foresounds
    - 1925 Backsounds
    - BehavioralDataExp3_v2.mat
        - BackSoundNum: (1-11) as shown in paper (1x1925 trials)
        - SNR: Signal-to-Noise-Ratio Value Between Fore and Back (1x1925 trials)
        - Resp: Behavioral Response (4 participants x 1925 trials) - 0 (incorrect response), 1 (correct response)

Supplemental Audio
- Audio 1
    - Example audio excerpts for each of the 33 background conditions for Experiment 1. All of the three-digit sequences are delivered at an SNR -9 dB. Exemplars for each of the 11 background sound conditions are organized in a directory corresponding to the experimental manipulation performed (OR, PR, SE). Sounds are numbered from 1 to 11 in rank order according to the perceptual accuracy, as shown in paper (Sound01.wav, Sound02.wav, etc.).
- Audio 2
    - Example audio excerpts for each of the background conditions for Experiment 2. The background conditions include the babble-8 and jackhammer backgrounds delivered at variable SNR and variable statistics (Spec, +Mar, +MPS, +Corr, and Orig). Sounds are organized in subdirectories corresponding to the SNR condition (-12, -9, -6, and -3 dB) and background name (EightSpeakerBabble or Jackhammer) and are labeled according to the summary statistics used to synthesize the background (SoundSpec.wav, SoundMar.wav, SoundMPS.wav, SoundCorr.wav, SoundOrig.wav).