Closed Projects

Vision-Based Mapping And Exploration for Home Robots

Researcher: Hedayat Vatankhah

Abstract. Considering the high complexity of home environments, a home robot cannot rely on sonar or 2D range sensors for effective navigation. Visual sensors provide a considerable amount of valuable information for robot navigation, and are cheaper than 2D and 3D laser sensors. The focus of this research is on vision based SLAM and a novel hybrid navigation method for a home robot. The goal is to enable a robot to be able to navigate in a dynamic, home like environment effectively.

Electronic Nose

Researcher: Masoud Nikoofar

Abstract. An electronic nose (e-nose) is a device that identifies the specific components of an odor and analyzes its chemical makeup to identify it. Electronic nose consists of a mechanism for chemical detection (an array of gas sensors), and a mechanism for pattern recognition. An odor has a special effect on each sensor that can change the resistor of sensors.
In other hands, we can detect kind of odor with resistor changing.

Sound Source Localization

Researcher: Seyed Mahdi Hosseini

Abstract. In this project real-time DSP based speech processing for embedded systems is focused to provide a robust sound source localizer for Sourena robot.

Robot Behavior learning using Dynamic Bayesian networks

Researcher: Abolfazl Nadi

Abstract. In this work we research on new structure for learning robot behavior based on dynamic Bayesian network, which outperforms the conventional Bayesian network in learning behaviors. For learning behaviors in presence of robot sensorís fault, a new fault detection and isolation framework based on Bayesian network is developed. This framework can tolerate all types of fault which take place in robot sensors. In this structure, raw sensor data are used to detect the faulty sensor.

Real-time 3D Face Recognition

Researcher: Mahmoud Rahat

Abstract. When robots are to recognize human in home environment they expose situation which have to recognize people from samples taken in uncontrolled circumferences and varying light condition. Here we present a novel face detection algorithm which uses voting strategy to improve efficiency of a boosted classifier for reduction of false positive samples. We apply a KNN classifier for recognition of localized faces. This algorithm learns new samples incrementally which enables robot to recognize new users afterwards.

Object Category Recognition

Researcher: Hadi Mahzarnia

Abstract. In this project the main objective is to model the human being behavior while facing a new object from the motion point of view. In other words, object categories is found by robot 2-D movements beside camera pan-tilt.

Object Recognition for Home Robots Inspired by the Primate Visual Cortex

Researcher: Mohsen Malmir

Abstract. Human can perform object recognition with high accuracy under a variety of object rotations and translations. The structure and function of the visual cortex has inspired many models for invariant object recognition. We propose a hierarchical model for object recognition based on the two well-known properties of the visual cortex neurons: invariant responses to stimulus transformations and redundancy reduction. We used the trace learning rule to provide the neurons in the model with invariant responses to object transformations.

Back to open research projects