Bayesian Action-Perception Computational Model for Cursive Letter Recognition and Production
Author Information
Author(s): Estelle Gilet, Julien Diard, Pierre Bessière
Primary Institution: Laboratoire d'Informatique de Grenoble, INRIA Rhône-Alpes, CNRS, Montbonnot, France
Hypothesis
The study investigates the interaction of perception and action representations involved in cursive letter recognition and production using a Bayesian model.
Conclusion
The Bayesian Action-Perception model effectively simulates various cognitive tasks related to reading and writing, demonstrating the influence of motor knowledge on perception.
Supporting Evidence
- The model successfully simulates six cognitive tasks related to reading and writing.
- Internal simulation of movements improves letter recognition accuracy.
- The model demonstrates the influence of motor knowledge on perception tasks.
- Recognition rates were high for complete letters but dropped for truncated letters.
Takeaway
This study shows how our brain uses both what we see and what we know about writing to recognize and produce letters.
Methodology
The study employs a Bayesian model to simulate cognitive tasks related to cursive letter recognition and production, using data collected from participants writing letters.
Potential Biases
The study may be biased by the limited number of participants and the specific writing styles they represent.
Limitations
The model's parameters were learned from a limited dataset, which may not generalize to all handwriting styles.
Participant Demographics
Four adults participated in the study, providing a database of handwritten letters.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website