Kiran Kumar Ravulakollu

Towards Computational Modelling of Neural Multimodal Integration Based on the Superior Colliculus Concept

Information processing and responding to sensory input with appropriate actions are among the most important capabilities of the brain and the brain has specific areas that deal with auditory or visual processing. The auditory information is sent first to the cochlea, then to the inferior colliculus area and then later to the auditory cortex where it is further processed so that then eyes, head or both can be turned towards an object or location in response. The visual information is processed in the retina, various subsequent nuclei and then the visual cortex before again actions will be performed. However, how is this information integrated and what is the effect of auditory and visual stimuli arriving at the same time or at different times? Which information is processed when and what are the responses for multimodal stimuli? Multimodal integration is first performed in the Superior Colliculus, located in a subcortical part of the midbrain. In this chapter we will focus on this first level of multimodal integration, outline various approaches of modelling the superior colliculus, and suggest a model of multimodal integration of visual and auditory information.

Stimuli Localization: An Integration Methodology Inspired by the Superior Colliculus for Audio and Visual Attention

Abstract Multiple Sensor fusion is important for human and animals to localize stimuli within an environment. Visual and audio sensors are often integrated to direct attention such as in human communication. The paper describes a methodology for designing and developing architecture to effectively localise simultaneous audio and visual stimuli by integrating both sensors. The Superior Colliculus (SC) inspired the architecture and it mimics the top and deep layers of the SC, as these layers are mainly responsible for visual and audio stimuli localization, respectively. The integration methodology described in this paper is evaluated against algorithmic-based methodology to determine effectiveness of the approach. Experimental results successfully demonstrate the key advantages in the integration, including (i) low-level multimodal stimuli localization and (ii) dimensionality reduction of the input-space without affecting stimuli strength and localization.

Analysis and evaluation of hybrid intrusion detection system models

Detecting intrusions is becoming an important and yet challenging task in information systems. This led to an increasing need for efficient methods of recognising intrusions in order to protect the systems. Existing models of intrusion detection systems (IDS) have produced significant performance but often possesses the inability for detecting multilevel classes of attacks coupled with high training time for classifiers. These drawbacks led to hybrid models that combine the various strengths of single classifiers at the same time avoiding their weaknesses for better performance. In this paper, a comparison of such hybrid models is carried out. The objective is to determine their performances and isolate their weaknesses. Thus, a research gap is established for more efficient intrusion detection models.

A novel methodology for exploring enhancement and depression phenomena in multisensory localization: A biologically inspired solution from Superior Colliculus

Localization is very essential for interaction when it comes to multisensory integration. Based on Superior Colliculus (SC) motivation, the audio and visual signal processing during the stimuli integration is investigated. A novel methodology is proposed using neural network architecture that can localize effectively, especially in integrating stimuli of varied intensities in lower order audio and visual signals. During the integration, cases arise where the SC is unable to localize the source due to simultaneous arrival of too weak or too strong stimuli, causing enhancement and depression phenomena. This phenomena arise when the SC is not able to localize the source based on the given stimuli intensities. This paper provides a dual layered neural network model that integrates visual and audio sensory stimuli and also drives a way to track the stimuli source. This behavior is applicable for guided robots that help humans to track or cooperate for tasks like personal assistance, route guidance and incident tracking applications.

Improving Robot-Human Communication by Integrating Visual Attention and Auditory Localization Using a Biologically Inspired Model of Superior Colliculus

Effective agent communication is always been an important modern area of research. This paper focuses on achieving greater precision in common by improving agent-human communication with the help of visual attention and auditory localization based on a simple model of the superior colliculus in the human brain system. The model receives individual visual and auditory sensory stimuli and combines them to generate an integrated stimulus predicting the location of the sound source. This integrated stimulus is used to generate a motor saccade of the visual system to attend to the sound. The computational model is based on a neural network approach with learning and is explored in experiments reflecting varied conditions to determine whether it mimes the performance of superior colliculus in auditory and visual stimuli integration. Finally with a evaluation strategy carried between unimodal and multimodal data, the efficiency of the computational model of Superior Colliculus is determined. Performance of the neural network based computational model has proven effective in terms of learning, the better performance of the integrated response over unimodal response and providing a realistic communication experience.