Multi-modal data fusion of Voice and EMG data for Robotic Control

Wearable electronic equipment is constantly evolving and is increasing the integration of humans with technology. Available in various forms, these flexible and bendable devices sense and can measure the physiological and muscular changes in the human body and may use those signals to machine control. The MYO gesture band, one such device, captures Electromyography data (EMG) using myoelectric signals and translates them to be used as input signals through some predefined gestures. Use of this device in a multi-modal environment will not only increase the possible types of work that can be accomplished with the help of such device, but it will also help in improving the accuracy of the tasks performed. This paper addresses the fusion of input modalities such as speech and myoelectric signals captured through a microphone and MYO band, respectively, to control a robotic arm. Experimental results obtained as well as their accuracies for performance analysis are also presented.

TLU-Net: A Deep Learning Approach for Automatic Steel Surface Defect Detection

Visual steel surface defect detection is an essential step in steel sheet manufacturing. Several machine learning-based automated visual inspection (AVI) methods have been studied in recent years. However, most steel manufacturing industries still use manual visual inspection due to training time and inaccuracies involved with AVI methods. Automatic steel defect detection methods could be useful in less expensive and faster quality control and feedback. But preparing the annotated training data for segmentation and classification could be a costly process. In this work, we propose to use the Transfer Learning-based U-Net (TLU-Net) framework for steel surface defect detection. We use a U-Net architecture as the base and explore two kinds of encoders: ResNet and DenseNet. We compare these nets' performance using random initialization and the pre-trained networks trained using the ImageNet data set. The experiments are performed using Severstal data. The results demonstrate that the transfer learning performs 5% (absolute) better than that of the random initialization in defect classification. We found that the transfer learning performs 26% (relative) better than that of the random initialization in defect segmentation. We also found the gain of transfer learning increases as the training data decreases, and the convergence rate with transfer learning is better than that of the random initialization.