Binary classification of spoken words with passive phononic metamaterials

TL;DR

This paper demonstrates that non-periodic phononic metamaterials can be designed and fabricated to perform binary classification of spoken words, offering a low-power alternative for machine learning tasks.

Contribution

The authors introduce a novel non-periodic phononic metamaterial design that can encode and distinguish spoken words, expanding the potential of phononic devices for AI applications.

Findings

Successfully classified spoken words using phononic metamaterials

Demonstrated low-power, passive computation with phononic structures

Showed feasibility of data-driven design for complex phononic functionalities

Abstract

Mitigating the energy requirements of artificial intelligence requires novel physical substrates for computation. Phononic metamaterials have a vanishingly low power dissipation and hence are a prime candidate for green, always-on computers. However, their use in machine learning applications has not been explored due to the complexity of their design process: Current phononic metamaterials are restricted to simple geometries (e.g. periodic, tapered), and hence do not possess sufficient expressivity to encode machine learning tasks. We design and fabricate a non-periodic phononic metamaterial, directly from data samples, that can distinguish between pairs of spoken words in the presence of a simple readout nonlinearity; hence demonstrating that phononic metamaterials are a viable avenue towards zero-power smart devices.