Machine Learning Approach to Decoding Insect Perception and Unlocking Perception-Based Control Methods for Bio-Intelligent Cyborg Insects

Morishima Lab , Department of Mechanical Engineering , Osaka university

Incoming...

.....

Machine Learning-Based Facial Expression Recognition Using Stretchable Strain Sensors For Rehabilitation System

Wahyudi Intelligent System Lab (WISE), International Islamic University Malaysia, Kuala Lumpur, Malaysia

Facial expression recognition (FER) enables computers or machines to identify human emotions. The FER system is used in self-driving cars, healthcare, and smart environments. Most of the facial expression systems are based on computer vision and image processing technologies. Computer vision technologies are expensive since they require massive memory and computation resources. It also depends on the environment changes. However, sensor technologies overcome all the limitations because they do not need massive memory or expensive computation resources and do not depend on environmental changes. This study aims to develop FER systems based on stretchable strain sensor data using machine learning in driving a rehabilitation system. Two different stretchable strain sensors (commercial and developed) are used to recognize four facial expressions (neutral, happy, sad, and disgusted). This study mainly focuses on the developed stretchable strain sensor, which is still developed in the laboratory and is not stable. So, the commercial stretchable strain sensor is used to analyze the developed sensor performance. The stretchable strain sensor data is time-series data with noise and high dimensionality. The datasets are normalized and aggregated to remove noise and high dimensionality. It is processed as an input to the machine learning model, and then it is compiled and fitted by five machine learning models, which are K- Nearest Neighbour (KNN), Decision Tree (DT), Support Vector Machine (SVM), Logistic Regression (LR) and Random Forest (RF) models. The training and testing results show that the RF model achieves higher accuracy than other machine learning models. The RF FER model is then implemented in the experimental hardware test of the facial expression-driven rehabilitation system. When facial expression neutral, happy, sad, and disgust emotion, the elbow rehabilitation system (ERS) motor speed is 60%,80%,0% and 30% of its full speed, respectively. The simulation results show that RF has achieved 96% and 90% accuracy, respectively, in recognizing the correct facial expression using the three commercial and four developed sensors, respectively. The offline hardware experimental test results show that the facial expression has driven rehabilitation system has successfully achieved 93% and 83% accuracy. The three commercial and four developed stretchable strain sensor data to drive the rehabilitation system’s speed according to the facial expression displayed. In the real-time experimental test on five subjects, the system has achieved an accuracy of 75% in regulating the rehabilitation system's speed based on the actual users' facial expressions. The proposed study's limitation is that the stretchable strain sensors are uncomfortable for data collection and tests. However, the experimental results have proven that the proposed methods can drive the rehabilitation machine to move according to the recognized facial expression. The proposed system can enhance the rehabilitation system's comfort and safety according to the patients need. It will help the patients to recover better and faster and eventually improves their quality of life.

Design and Performance Evaluation of a Campus Network Based on Different Routing Protocols

International Islamic University Malaysia

A Campus Area Network (CAN) can be characterized as a Local Area Network (LAN). Generally, it is a network which connects various entities of an office e.g. various infrastructural entities of a university campus. In this paper, it is designed a new CAN topology and analyzed the performance of CAN by considering Routing Information Protocol (RIP), Enhanced Interior Gateway Routing Protocol (EIGRP), and Open Shortest Path First (OSPF) routing protocols. The simulation results were analyzed with a performance comparison among these three routing protocols. It is found that the proposed design model reduces the cost from bandwidth perspective and also provides values for various network parameters which lead to better Quality of Service (QoS) in the CAN.