Seminar
Sensor Based Complex Activity Recognition for Multi-resident in Smart Home 17388 days ago on 10-08-2017 10:00:00am, Seminar Room C, Block C, Faculty of Computer Science and Info. Tech., UPM
In recent times, we have seen the emergence of smart objects connecting each other to the physical world. The convergence of sensors and actuators would react automatically to the users within an environment. The sensors are hidden from the user so they become part of the environment, neither require the user to explicitly interact nor wear the devices. Recent research trends evolve not only on the devices and equipment that have "brain" to talk to each other, but to infer on human activities in the environment. The development of ambient intelligent and smart environment is sharing their common goal in supporting and assisting people in activities of daily living. Smart home in particular, promote the longevity of elderly to stay at home independently, increase the quality of life of home dwellers in term of security and safety. This talk highlights multi resident activity recognition in smart home.
Speaker's Biodata:Raihani Mohamed received her B.Sc (Management Information Systems) from Universiti Islam Antarabangsa Malaysia and obtained her M.Sc (Information Technology) from Universiti Teknologi MARA, Malaysia. Her research interests are in smart home systems, ambient assisted living environment, machine learning technique and human activity recognition. Her current research is on recognition of activity of multi-residents in smart home.
Recognizing of Physical Activities using Combination of Spectral Frequency and Statistical Descriptors Features 10th of August, 10.45-11:30, Seminar Room C, Block C, Faculty of Computer Science and Info. Tech., UPM
Prior knowledge in pervasive computing recently garnered a lot of attention due to its high demand in various application domains. Human activity recognition is considered as the applications that are widely explored providing valuable information to human. Accelerometer sensor-based approach is utilized as devices to undergo the work on human activity recognition due to their small size and it is available in almost modern smartphones. However, the existence of high inter-class similarities among the classes tends to degrade the recognition performance. To differentiate between stationary and locomotion activities, combination of spectral frequency and statistical descriptors can be used to improve activity recognition. The unwanted information is filtered using Fourier Transform before the features are extracted. These features are later classified using random forest ensemble classifier. Find out how.
Speaker's Biodata:Muhammad Noorazlan Shah Zainudin received his B. Sc (Computer Science) from Universiti Teknologi Malaysia and obtained his M.Sc (Computer Science) from Universiti Teknologi Malaysia. His research interests are in human activity recognition, pattern recognition, optimization algorithm and machine learning application. His current research is on recognition of human activity of acceleration sensor.
