DeepSpectrumLite: A Power-Efficient Transfer Learning Framework for Embedded Speech and Audio Processing From Decentralized Data

Keyphrases

• Amount of Training 50%
• Audio 50%
• Audio Processing 100%
• Audio Recognition 50%
• Audio Signal 50%
• Audio Signal Processing 50%
• Classification Task 50%
• Command Line Interface 50%
• Complex Architecture 50%
• Computational Power 50%
• Consumer-grade 50%
• Convolutional Neural Network 50%
• Cough 50%
• COVID-19 100%
• Data Power 50%
• Data Upload 50%
• Decentralized Data 100%
• Deep Neural 50%
• DenseNet121 50%
• Efficient Transfer Learning 100%
• Embedded Audio 50%
• Embedded Devices 50%
• Further Processing 50%
• Learning Framework 100%
• Mel-spectrogram 50%
• Moto 50%
• Motorola 50%
• Music Emotion Recognition 50%
• On the Fly 50%
• On-device 50%
• Power Efficient 100%
• Pre-trained Convolutional Neural Networks 50%
• Processing System 50%
• Publicly Available 50%
• Raw Audio 50%
• Real-world Application 50%
• Smartphone 50%
• Snore Sound Classification 50%
• Social Signal Processing 50%
• Speech Analysis 50%
• Speech Emotion Recognition 50%
• Speech Processing 100%
• Speech Recognition 50%
• Target Classification 50%
• Training Data 50%
• Training Parameters 50%
• Transfer Learning 50%
• Transfer Learning Model 50%

Computer Science

• Audio Recognition 33%
• Audio Signal Processing 100%
• Classification Task 33%
• Command Line Interface 33%
• Computational Power 33%
• Convolutional Neural Network 66%
• Embedded Device 33%
• Learning Approach 33%
• Learning Framework 100%
• Open Source 33%
• Power Efficient 100%
• Processing System 33%
• Social Recognition 33%
• Speech Analysis 33%
• Speech Processing 100%
• Speech Recognition 33%
• Training Data 33%
• Transfer Learning 100%
• World Application 33%