An Ultra-Low-Power Keyword-Spotting Processor with Trainable MFCC-CNN Framework and Multiplication-Efficient Acceleration

Key Points

• Demonstrates an efficient keyword-spotting process, optimizing power usage dramatically.
• Achieves up to 97% accuracy in keyword recognition while using minimal power resources.
• Utilizes a unique architecture based on a convolutional neural network and mel-frequency cepstral coefficients.
• Highlights the potential for integrating advanced keyword recognition in portable, battery-operated devices.