Recognizing configurations is the interaction between perception and memory 3

Recognizing configurations is the interaction between

This preview shows page 21 - 24 out of 108 pages.

Recognizing configurations is the interaction between perception and memory 3) An emerging percept has to mingle with a memory trace of an experience by some function The Three main theories of pattern recognition 1) Template Matching Theory - Pattern recognition would be simple if all patterns were the same, but patterns are not all the same o I.e. there is not just one way a chair appears How are we able to identify them as the same objects? - Theory suggests that we compare the configuration of the current sensory input with a standard configuration we have in memory (prototype or template) o A prototype is an average representation of the object in memory o If the match is good enough, we recognize the object Hintzman Multiple-trace Memory Model:
Image of page 21
- A probe from primary memory (current experience of the stimulus) activates secondary memory (previous experience of similar stimulus) where memory traces are activated to the extent that they are similar to the probe - Contributions from the activated memory traces return an echo to primary memory Probe – A snapshot of information in primary memory that can activate memory traces in secondary memory Echo – When a Probe goes out from primary memory to secondary memory, memory traces are activated to the extent that they are similar to the probe; made up of all contributions from memory traces Strengths – A simple economical explanation Weaknesses – It would be quite inefficient to build an inventory of templates to classify all the different varieties (too many templates!) What entails a match is not well defined at all. 2) Feature detection theory - We decompose objects in features and we identify them in such a way. - Feature – A very simple element that can appear in a combination with other visual elements across a variety of patterns - So it is the ultimate combination that allows recognition Selfridge’s Pandomenium model 1) Bottom level contains data demons , or images in which pattern of feature is represented 2) Next level contains cognitive demons ready to detect a particular pattern, if detected the demon shouts 3) Top level contains the decision demons , which selects the cognitive demon that is best represented from the second level a. Determine who shouts the loudest ** The Greatest similarity = The louder a cognitive demon will shout Importance of features - Similar to the primary visual cortex which has special cells and areas for processing simple visual features Implication of Pandemonium - Parallel processing o Perform many computation at once rather than in a serial manner o Features and patterns are being detected in parallel o Explains why we can recognize object quickly and efficiently o Each group of demons are looking for a particular aspect - Perception is a problem to solve o The solution relies on a group – the cognitive demons o This is adaptive The match does not have to be an exact match, rather one demon just has to be the loudest
Image of page 22
Problems - Feature detection involves some sort of list of defining features… -
Image of page 23
Image of page 24

  • Left Quote Icon

    Student Picture

  • Left Quote Icon

    Student Picture

  • Left Quote Icon

    Student Picture