Optimal features for auditory categorization
| Autor: | Xiaoqin Wang, Srivatsun Sadagopan, Shi Tong Liu, Pilar Montes-Lourido |
|---|---|
| Jazyk: | angličtina |
| Rok vydání: | 2018 |
| Předmět: |
Male
0301 basic medicine Auditory perception Sound Spectrography Computer science Science Speech recognition Guinea Pigs High variability General Physics and Astronomy 02 engineering and technology Auditory cortex Article General Biochemistry Genetics and Molecular Biology Membrane Potentials Stereotaxic Techniques 03 medical and health sciences Feature (machine learning) otorhinolaryngologic diseases Animals Humans Layer (object-oriented design) Set (psychology) Greedy algorithm lcsh:Science Auditory Cortex Neurons Multidisciplinary Callithrix General Chemistry Function (mathematics) 021001 nanoscience & nanotechnology Electrodes Implanted Identification (information) Sound ComputingMethodologies_PATTERNRECOGNITION 030104 developmental biology Acoustic Stimulation Categorization Stereotaxic technique Auditory Perception Female lcsh:Q Vocalization Animal 0210 nano-technology |
| Zdroj: | Nature Communications, Vol 10, Iss 1, Pp 1-14 (2019) Nature Communications |
| DOI: | 10.1101/411611 |
| Popis: | Humans and vocal animals use vocalizations to communicate with members of their species. A necessary function of auditory perception is to generalize across the high variability inherent in vocalization production and classify them into behaviorally distinct categories (‘words’ or ‘call types’). Here, we demonstrate that detecting mid-level features in calls achieves production-invariant classification. Starting from randomly chosen marmoset call features, we use a greedy search algorithm to determine the most informative and least redundant features necessary for call classification. High classification performance is achieved using only 10–20 features per call type. Predictions of tuning properties of putative feature-selective neurons accurately match some observed auditory cortical responses. This feature-based approach also succeeds for call categorization in other species, and for other complex classification tasks such as caller identification. Our results suggest that high-level neural representations of sounds are based on task-dependent features optimized for specific computational goals. Vocalizations such as speech or animal calls have high variability in production. Here, the authors report that a few mid-level acoustic features provide sufficient information to generalize across this variability and classify vocalization types and auditory cortical neurons exhibit tuning to these features. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|