How to Build an ML-Powered Doorbell Notifier Using TensorFlow Lite and Arm Cortex-M33

Learn how to create an ML-powered doorbell notification system in this technical talk from the tinyML EMEA conference. Explore the development of an audio classification tinyML system built on an Arm Cortex-M33 based Realtek RTL8721DM SoC. Discover techniques for working with audio datasets and implementing transfer learning to train TensorFlow Lite models using TensorFlow's signal processing and Keras APIs. Master the process of porting feature extraction pipelines to SoC using CMSIS-DSP and running ML interfaces with TensorFlow Lite for Microcontrollers, enhanced by CMSIS-NN accelerated computation kernels. Understand how to optimize Wi-Fi connectivity by triggering it only when the model detects relevant audio sounds. Gain insights into the Realtek RTL8721DM SoC's capabilities, including its compute resources that enable exploration of complex model architectures for various tinyML audio classification applications. Access the complete project code on GitHub to implement and expand upon the demonstrated concepts, from preprocessing and spectrogram generation to MEMS microphone input handling and Arduino integration.