Ecological origins of perceptual grouping principles in the auditory system - stimulus examples

Wiktor Młynarski & Josh H. McDermott

Department of Brain and Cognitive Sciences, MIT

mlynar@mit.edu, jhm@mit.edu


This site accompanies the manuscript "Ecological origins of perceptual grouping principles in the auditory system".
Click on stimulus cochleagrams to hear sounds. For best results please use headphones.

Sound data information:
All results were derived from a corpus of sounds generated by individual physical sources. The corpus was created by merging corpora of recordings of individual talkers and musical instruments in equal proportion. Speech sounds were taken from the TIMIT database [1], and included voices of male and female speakers speaking sentences in English. Solo instrument sounds were taken from the RWC Music Database [2].

[1] Garofolo, J. S. & Consortium, L. D. TIMIT: Acoustic-phonetic continuous speech corpus. (Linguistic Data Consortium, 1993).
[2] Goto M., Hashiguchi H., Nishimura T., and Oka R. RWC Music Database: Music Genre Database and Musical Instrument Sound Database. In proceedings ISMIR 2003, October 2003.


Features learned from natural sound statistics




Co-occurring feature mixtures




Non-co-occurring feature mixtures






Spectrotemporal cue 1

Grouping feature mixtures (low cue value)

All 4 cues have low values

Non-grouping feature mixtures (high cue value)

Remaining 3 cues have low values

Spectrotemporal cue 2

Grouping feature mixtures (low cue value)

All 4 cues have low values

Non-grouping feature mixtures (high cue value)

Remaining 3 cues have low values

Modulation cue 1

Grouping feature mixtures (low cue value)

All 4 cues have low values

Non-grouping feature mixtures (high cue value)

Remaining 3 cues have low values

Modulation cue 2

Grouping feature mixtures (low cue value)

All 4 cues have low values

Non-grouping feature mixtures (high cue value)

Remaining 3 cues have low values




Full discriminative model decisions - feature pairs

Sound pairs classified as a single source

Sound pairs classified as different sources


Full discriminative model decisions - artificial sound ("blob") pairs

Sound pairs classified as a single source

Sound pairs classified as different sources


Full discriminative model decisions - apertured speech pairs

Sound pairs classified as a single source

Sound pairs classified as different sources





Feature sequences

Reference sequence

Mixture with a co-occurring sequence

Mixture with a non-co-occurring sequence