Proposing a Recognition System of Gestures Using MobilenetV2 Combining Single Shot Detector Network for Smart-Home Applications

Huu, Phat Nguyen; Thu, Huong Nguyen; Minh, Quang Tran

doi:10.1155/2021/6610461

Cited by 10 publications

(4 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Zhao et al [ 20 ] focused on the ability of neural network in feature extraction so that the feature extraction ability obtained by neural network on ImageNet dataset was applied to the extraction of eye features, which solved the problem of small eye dataset and enabled the optimal feature vector to be obtained with less data, resulting in high eye recognition rate. Huu et al [ 21 ] proposed a gesture recognition system based on MobilenetV2, which can achieve good performance on small-scale data sets by reusing the features learned in the network through dense connection.…”

Section: Literature Reviewmentioning

confidence: 99%

Evaluation Technology of Classroom Students’ Learning State Based on Deep Learning

Chen

2021

Computational Intelligence and Neuroscience

View full text Add to dashboard Cite

Facial features are an effective representation of students’ fatigue state, and the eye is more closely related to fatigue state. However, there are three main problems in the existing research: (1) the positioning of the eye is vulnerable to the external environment; (2) the ocular features need to be artificially defined and extracted for state judgment; and (3) although the student fatigue state detection based on convolutional neural network has a high accuracy, it is difficult to apply in the terminal side in real time. In view of the above problems, a method of student fatigue state judgment is proposed which combines face detection and lightweight depth learning technology. First, the AdaBoost algorithm is used to detect the human face from the input images, and the images marked with human face regions are saved to the local folder, which is used as the sample dataset of the open-close judgment part. Second, a novel reconstructed pyramid structure is proposed to improve the MobileNetV2-SSD to improve the accuracy of target detection. Then, the feature enhancement suppression mechanism based on SE-Net module is introduced to effectively improve the feature expression ability. The final experimental results show that, compared with the current commonly used target detection network, the proposed method has better classification ability for eye state and is improved in real-time performance and accuracy.

show abstract

Section: Literature Reviewmentioning

confidence: 99%

Evaluation Technology of Classroom Students’ Learning State Based on Deep Learning

Chen

2021

Computational Intelligence and Neuroscience

View full text Add to dashboard Cite

show abstract

“…Huu, P . N [25] proposed a system for recognizing gestures and actions in smart homes. They used actions such as walking, sitting, backing, putting on shoes, waving, falling, smoking, infant crawling, standing, reading, and typing for recognition.…”

Section: Introductionmentioning

confidence: 99%

Smart home service terminal design for elderly families integrating the KANO model and perceptual engineering

Chen,

Tang,

Yan

2023

3C Tecnología

View full text Add to dashboard Cite

In recent years, smart homes have gradually come into our lives and have brought many positive impacts to our lives. However, in specifically targeted application design planning, the designer community is not always able to consider and analyze every factor. This paper proposes an integrated nonlinear design that incorporates the KANO model as well as a mathematical model of sensible engineering coupled with design science. The application of different design solutions in different types of households is evaluated after dividing the different types of living of elderly households into specific situations. The results show that for elderly households of type 1 versus 4, scenario 2 generally has more accurate application feasibility compared to scenario 1. The maximum increase in application accuracy for Scenario 2 compared to Scenario 1 was 6.28%. However, the frequency of use decreased by 3.09%. And for elderly households of type 2 and 3, which tend to live alone, the feasibility of application of scenario 2 is similar or even worse than that of scenario 1. The improved Scenario 3 both have higher application feasibility than Scenarios 1 and 2 and has a more user-friendly visual aid to understand the design, which helps the elderly group to better use the smart home service terminal system.

show abstract

“…Other researchers developed different SLTs such as software-based platform [36], MobilenetV2based gestures recognition system [37], and tablet-based hearing aid [38]. An open-source software framework developed in [36] presents a development environment for building of augmentative and alternative communication models that include communication aids for the disabled community.…”

mentioning

confidence: 99%

“…An open-source software framework developed in [36] presents a development environment for building of augmentative and alternative communication models that include communication aids for the disabled community. A MobilenetV2-based gesture recognition system was developed in [37]. Although, the system is specifically meant for smart home applications, it can as well be deployed for gestures translation.…”

mentioning

confidence: 99%

Development of a PC-based sign language translator

Amusa

Olanipekun

Erinosho

et al. 2023

IJ-ICT

View full text Add to dashboard Cite

While a hearing-impaired individual depends on sign language and gestures, non-hearing-impaired person uses verbal language. Thus, there is need for means of arbitration to forestall situation when a non-hearing-impaired individual who does not understand the sign language wants to communicate with a hearing-impaired person. This paper is concerned with the development of a PC-based sign language translator to facilitate effective communication between hearing-impaired and non-hearing-impaired persons. Database of hand gestures in American sign language (ASL) is created using Python scripts. TensorFlow (TF) is used in the creation of a pipeline configuration model for machine learning of annotated images of gestures in the database with the real time gestures. The implementation is done in Python software environment and it runs on a PC equipped with a web camera to capture real time gestures for comparison and interpretations. The developed sign language translator is able to translate ASL/gestures to written texts along with corresponding audio renderings at an average duration of about one second. In addition, the translator is able to match real time gestures with the equivalent gesture images stored in the database even at 44% similarity.

show abstract

Proposing a Recognition System of Gestures Using MobilenetV2 Combining Single Shot Detector Network for Smart-Home Applications

Cited by 10 publications

References 23 publications

Evaluation Technology of Classroom Students’ Learning State Based on Deep Learning

Evaluation Technology of Classroom Students’ Learning State Based on Deep Learning

Smart home service terminal design for elderly families integrating the KANO model and perceptual engineering

Development of a PC-based sign language translator

Contact Info

Product

Resources

About