Machine pattern recognition mirrors Gestalt grouping via feature vectors and similarity metrics

Evidence

The theory and techniques of pattern recognition have developed through research into machine vision and machine hearing. In pattern recognition, information captured from the real world (for example, a visual or auditory signal) is pre-processed and transformed into a vector of relevant features. The acquired feature vector is then compared to a set of known feature vectors using a suitable algorithm, with the goal of recognizing (classifying) the unknown pattern. Pattern recognition often requires the formation of quantitative metrics for distance or similarity (compare with the principles of grouping). — Ville Pulkki, Matti Karjalainen, Communication Acoustics- An…, 1327

Categorization

Category: Connection Points

• XREF: Direct bridge from perceptual grouping principles to machine-learning classification — the foundation of modern speech recognition and audio tagging.

Related

• Machine hearing mirrors human perception by comparing feature vectors against known patterns (similarity 0.90)
• Gestalt grouping principles predict relative pitch perception performance (similarity 0.83)
• Auditory Gestalt perception mirrors vision through spectral-pitch contours (similarity 0.81)
• Gestalt grouping principles apply to hearing as well as vision (similarity 0.81)
• Auditory grouping by pitch proximity can override melodic continuation perception (similarity 0.78)

Broader Connections

• Gestalt grouping principles predict relative pitch perception performance #exemplifies (0.85) — Dijk, Pim van, Physiology, Psychoacoustics and…, 5270
• Small within-pattern intervals improve transposed pattern comparison via featural grouping #expands (0.82) — Dijk, Pim van, Physiology, Psychoacoustics and…, 5263
• Binaural illusions reveal the brain imposes Gestalt grouping on incoming pitches #exemplifies (0.80) — David M. Howard, Jamie Angus, Acoustics and Psychoacoustics, 1620
• Neural networks effectively represent words as points in vector space #exemplifies (0.78)