Efficient Word-Level Sign Language Recognition Using Quantized Spatiotemporal Deep Learning for Low-Power Microcontrollers

Keyphrases

• Low-power Microcontroller 100%
• Conv1D 100%
• Spatiotemporal Deep Learning 100%
• Word-level Sign Language Recognition 100%
• INT8 100%
• Microcontroller 66%
• TinyML 66%
• Sign Language Recognition 66%
• Memory Processing 33%
• Enhanced Stability 33%
• Energy Conservation 33%
• Architectural Design 33%
• Recognition Model 33%
• Size Reduction 33%
• Human-computer Interaction 33%
• Real-time Performance 33%
• Privacy-preserving 33%
• MobileNet 33%
• Computation Constraint 33%
• Evaluation Scale 33%
• TensorFlow Lite 33%
• Ultralight 33%
• Memory Constraints 33%
• Art Architecture 33%
• Edge Devices 33%
• Cross-dataset Generalization 33%
• Temporal Processing 33%
• CMSIS-NN 33%
• Class Constraint 33%
• Spatiotemporal Architecture 33%
• Stability Operator 33%
• Deployment pipeline 33%
• CNN+RNN 33%
• Large Vocabulary 33%
• Spatial Computing 33%
• Quantization Accuracy 33%
• Integer-only 33%
• Quantization-aware Training 33%
• Embedded Constraints 33%
• Isolated Word 33%
• Attention Operator 33%
• Recurrent Operators 33%
• INT8 Quantization 33%
• Recognition Study 33%

Computer Science

• Microcontroller 100%
• Deep Learning Method 100%
• Sign Language Recognition 100%
• Privacy Preserving 20%
• Real Time Performance 20%
• Long Short-Term Memory Network 20%
• Recurrent Neural Network 20%
• Convolutional Neural Network 20%
• Architectural Design 20%
• Mobilenet 20%
• TensorFlow 20%
• Server-Class 20%