Brain Architecture for Visual Object Identification

Brain Architecture for Visual Object Identification

Gustavo Torres, Karina Jaime, Félix Ramos
DOI: 10.4018/jcini.2013010104
(Individual Articles)
No Current Special Offers


Visual memory identification is a key cognitive process for intelligent virtual agents living on virtual environments. This process allows the virtual agents to develop an internal representation of the environment for the posterior production of intelligent responses. There are many architectures based on memory modules for environment visual elements identification, as if they were invariant, this way of processing a visual scene is different from the one that real humans use. This document presents the description of a visual memory identification model based on current neuroscience state of art. Furthermore; the proposed model considers memory as a system that treats information in three stages: to encode, store and retrieve acquired knowledge about the environment. On the other hand, the authors validate the implementation of their approach with two identification tasks: when the stimulus is known and when it is unknown. Actually, this work is part of a proposal for a cognitive architecture that will let the authors create virtual agents with more credible human behaviors.
Article Preview


Visual information is a key component of intelligent behavior. Through vision, humans perceive their surroundings. Visual information is rich in detail, which allows humans to remain at a safe distance from the stimuli sources. For that reason, a large amount of human intelligent behavior relies on visual stimulation. In addition, memory is an important process of human behavior because it enables the brain to encode, store and retrieve acquired knowledge about the environment (Kandel & Kufermann, 2000). Memory (together with other cognitive functions) creates predictive scenarios, and generates thoughts, decisions, and responses. As will be explained below, visual memory is involved in object identification by discriminating new data from stored data.

One of the main abilities of human visual system is to associate the input stimuli with the stored knowledge about the world; namely, this association allows humans to understand the meaning of the visual scene. In cognitive architectures for intelligent agents, the visual system has become an important process. As an example of this, Wang (2007) defines perception as the cognitive processes at subconscious function that detects, make relations and interpretations, and searches internal information in the mind. Later he explained that, among other things, perception determines personality and will be an important part of cognitive computers.

For its study, memory has been divided into two systems: short-term memory and long-term memory. When past information has not been stored, it has been retained in short-term memory. Retained past events are stored in long-term memory (Atkinson & Shiffrin, 1968). The long-term memory system holds two kinds of information: declarative or explicit data and non-declarative or implicit data (Kandel & Kufermann, 2000). Declarative memory does the effort of recalling events and facts. Non-declarative memory retains information regarding well-structured and effortless processes: habits, reflexes, motor and emotional responses (Milner, Squire, & Kandel, 1998). The short-term memory and long-term memory systems can be found in different brain processes: sensory, cognitive and metacognitive brain operations.

It is recognized that computer vision supports the visual modeling in different cognitive architectures. Computational vision sees perception as a set of rules related to specific environmental information extraction from certain features of the stimuli (Dickinson, 1999). Within this framework, the visual system is made up of two modules: a device to capture and split information about the world, converting it into a signal, and a center that synthesizes and processes these signals in a significant internal image. For cognitive architectures, images in visual system are related to a kind of stored memory. Even though, artificial cognitive architectures (ACA's) usually rely on different forms of memory systems, none have developed visual memory. This is an important point, because delineating this system assumes the dynamic participation of several modules of the cognitive architecture. Stressing communication between these modules is as important as module construction to deal with all forms of visual information. Existing ACA's rely on long-term and short-term visual memory systems, just as neuropsychology assumes that human brains do. Visual memory depends on object features given to the memory modules in different ways, as we will describe below for the most important ACA's.

The Soar architecture takes the current visual image as part of the features of the current state in which intelligent agent is involved; namely, some attributes of the current state are provided by the visual system (Laird, Newell & Rosenbloom, 1987). The visual system adds information to be manipulated in working memory. Soar assumes that visual system only contributes to the representation of the agents’ state. Soar's (Lathrop & Laird, 2007) memory system includes a declarative or short-term memory and a procedural or long-term memory. Short-term memory represents the agent's current state. Some symbols might represent an object that matches the current perception or a past memory activation. Interaction with long-term memory involves changing the objects' features in short-term memory.

Complete Article List

Search this Journal:
Volume 17: 1 Issue (2023)
Volume 16: 1 Issue (2022)
Volume 15: 4 Issues (2021)
Volume 14: 4 Issues (2020)
Volume 13: 4 Issues (2019)
Volume 12: 4 Issues (2018)
Volume 11: 4 Issues (2017)
Volume 10: 4 Issues (2016)
Volume 9: 4 Issues (2015)
Volume 8: 4 Issues (2014)
Volume 7: 4 Issues (2013)
Volume 6: 4 Issues (2012)
Volume 5: 4 Issues (2011)
Volume 4: 4 Issues (2010)
Volume 3: 4 Issues (2009)
Volume 2: 4 Issues (2008)
Volume 1: 4 Issues (2007)
View Complete Journal Contents Listing