Exploring Sign Language Detection on Smartphones: A Systematic Review of Machine and Deep Learning Approaches

Table 5

Techniques of sign language recognition using smartphones.

Study	Year	Techniques	Evaluation metric
[153]	2023	DeepVision transformers	Accuracy, precision
[154]	2023	8-Layer CNN	Accuracy
[155]	2023	K-nearest neighbors (KNN)	Accuracy
[150]	2023	Deep learning (DL) combined with CNN and RNN	Accuracy
[147]	2023	DNN	Accuracy with [email protected]
[146]	2022	CNN	Accuracy
[144]	2022	SVM	Accuracy
[143]	2022	Inaudible acoustic signal to estimate channel information and capture the sign language in real time	Accuracy
[156]	2022	CNN	Accuracy
[157]	2022	CNN, DCGAN	Accuracy
[141]	2021	SVM	Accuracy, precision, recall, F1 score
[158]	2021	CNN	Accuracy
[159]	2021	3DCNN	Accuracy
[160]	2021	CNN, RNN	Accuracy
[137]	2020	ISL parser, Hamburg notation system, signing gesture markup language, 3D avatar	BLEU score, accuracy
[138]	2020	CNN	Word recognition rate
[127]	2020	Long short-term memory (LSTM)	Accuracy
[128]	2020	AutoML, transfer learning	Precision, recall, F1 score, accuracy
[129]	2020	MobileNet and ResNet	Accuracy
[133]	2020	MobileNet	Accuracy
[132]	2020	MobileNet-V3	Accuracy
[120]	2020	Artificial neural networks (ANNs)	Accuracy
[102]	2019	State-of-the-art pose estimation method	Accuracy
[110]	2019	CNN	Accuracy
[105]	2019	Simple classification algorithms from machine learning	Accuracy
[103]	2019	SVM	Accuracy, precision, recall, F measure
[104]	2019	SVM	Accuracy, precision, recall, specificity, F1 measure
[109]	2019	Elliptical Fourier descriptor and LSTM	Training time, testing time, accuracy
[119]	2019	AdaBoost, multilayer perceptron, Naïve Bayes, random forest, SVM, dynamic feature selection and voting	Accuracy
[91]	2019	CNN, LSTM, and connectionist temporal classification (CTC)	Accuracy, WER
[101]	2019	MIT invertor	Accuracy
[90]	2019	LSTM and CTC	Accuracy, WER
[87]	2019	OpenPose, hidden Markov model	Accuracy
[115]	2019	Gesture recognition algorithm of talking hands	Accuracy
[74]	2018	Flex sensor with Arduino	Accuracy
[70]	2018	CNN	Accuracy
[84]	2018	CNN	Accuracy, recognition time
[82]	2018	Naïve Bayes, multilayer perceptron (MLP)	Accuracy, F1 score
[75]	2018	KNN	Accuracy, recognition time
[79]	2018	ANN	Word matching score (WMS)
[83]	2018	ANN, minimum distance classifier	WMS
[60]	2017	Neural network	N.A
[62]	2017	Principle component analysis	Accuracy
[66]	2017	Word matching score (WMS) and ANN	WMS
[56]	2017	SVM, Naïve Bayes, random forest	Accuracy, F1 score
[57]	2017	Binarized neural network, LSTM	Detection ration (DR), reliability ration (RR), WER
[2]	2017	KNN, SVM linear, radial basis function SVM, random forest	F measure, ROC, accuracy
[67]	2017	Discrete-time warping	Accuracy
[61]	2017	Arduino	N.A
[50]	2016	SVM	Accuracy
[49]	2016	Backpropagation neural network	Accuracy
[45]	2016	Dynamic time warping	Recognition time, extensibility, recognition time (accuracy)
[52]	2016	Euclidean, normalized Euclidian, and Mahalanobis distance	WMS
[51]	2016	Optical character recognition, Microsoft Arabic Toolkit Service (ATKS), named entity recognizer (NER)	Recognition time, usability
[41]	2015	Neural networks (NNs) with log-sigmoid, NN with symmetric Elliott, and SVM	Accuracy, classification time, memory usage, battery consumption
[42]	2015	Microcontroller	Accuracy
[39]	2015	Flex sensors, inertial sensors	Sensitivity, accuracy
[37]	2015	KNN classification. The time needed by the system to recognize a single sign is between 6 frames per second (FPS) and 20 FPS.	Accuracy
[40]	2015	Arduino	Accuracy, error rate
[28]	2014	Recognition algorithm using histogram of oriented gradients (HOG)	Recognition rate, processing time
[33]	2014	Principle component analysis (PCA) for feature extraction and Euclidean distance for classification	Accuracy
[26]	2012	Sign modeling language (SML), animation engine	N.A