This work present a novel in-filter computing framework that can be used for designing ultra-light acoustic classifiers for use in smart internet-of-things (IoTs). Unlike a conventional acoustic pattern recognizer, where the feature extraction and classification are designed independently, the proposed architecture integrates the convolution and nonlinear filtering operations directly into the kernels of a Support Vector Machine (SVM)…

The result of this integration is a template-based SVM whose memory and computational footprint (training and inference) is light enough to be implemented on an FPGA-based IoT platform.

The tinyRadar project proposes a novel tinyML-based single-chip radar solution for on-edge sensing and detection of human activity. Edge computing within a small form factor solves the issue of data theft and privacy concerns as radar provides point cloud information. 

It can operate in adverse environmental conditions like fog, dust, and low light. We used the Texas Instruments IWR6843 millimeter wave (mmWave) radar board to implement signal processing chain on DSP C67x and Convolutional Neural Network (CNN) on Cortex®-R4F MCU for real-time inference of human activity.

A dataset for four different human activities generalized over six subjects was collected to train the 8-bit quantized CNN model. The real-time inference engine implemented on Cortex®-R4F using CMSIS-NN framework has a model size of 1.44KB, gives the classification result after every 120ms, and has an overall subject-independent accuracy of 96.43%.

 

Neuromorphic Trainable Analog Integrated Chip