Search

Vision-based Continuous Sign Language Recognition (CSLR) is a challenging and weakly supervised task aimed at segmenting sign language from weakly annotated image stream sequences for recognition. Compared with Isolated Sign Language Recognition (ISLR), the biggest challenge of this work is that the image stream sequences have ambiguous time boundaries. Recent CSLR works have shown that the visual-level sign language recognition task focuses on image stream feature extraction and feature alignment, and overfitting is the most critical problem in the CSLR training process. After investigating the advanced CSLR models in recent years, we have identified that the key to this study is the adequate training of the feature extractor. Therefore, this paper proposes a CSLR model with Multi-state Feature Optimization (MFO), which is based on Fully Convolutional Network (FCN) and Connectionist Temporal Classification (CTC). The MFO mechanism supervises the multiple states of each Sign Gloss in the modeling process and provides more refined labels for training the CTC decoder, which can effectively solve the overfitting problem caused by training, while also significantly reducing the training cost in time. We validate the MFO method on the popular CSLR dataset and demonstrate that the model has better performance. [ABSTRACT FROM AUTHOR]

Thin-walled micro parts are widely used in many fields, such as aviation, aerospace, precision engineering, and even micro-molds. Compared with other micro-fabrication technology, micro-milling can be considered as one of the most efficient 3D fabrication techniques. However, due to the vulnerable stiffness of thin-walled micro parts and micro tools, regenerative chatter is prone to occur during the micro-milling process. Severe chatter can lead to large fluctuations in cutting forces, resulting in deterioration of surface roughness/integrity and decreasing in tool life. Therefore, an in-process chatter detection strategy based on the feature extraction of micro-milling forces is proposed in this paper. Micro-milling force signals are measured and processed by advanced algorithms developed. After variational mode decomposition (VMD) for each group of cutting forces, the optimal intrinsic mode function (IMF) is adopted based on the Laplacian score (LS). The multi-scale permutation entropy (MSPE) values of the optimal IMF of each group are calculated, wherein the scale factor s , the embedding dimension m , and the delay time t are obtained by the genetic algorithm (GA). The obtained MSPE values are used as input vectors to train the support vector machine (SVM) classifier, which is used to monitor the stability states of the micro-milling processes. The effectiveness of the proposed strategy is compared with the other classic chatter detection methods and indicators in the micro-milling experiments of thin-walled parts. The results show that the method resulted from the strategy can extract cutting force features effectively and in-process detect chatters accurately in micro-milling thin-walled parts processes. [ABSTRACT FROM AUTHOR]