We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

eess

Change to browse by:

References & Citations

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Sound

Title: Ultra-Low Power Keyword Spotting at the Edge

Abstract: Keyword spotting (KWS) has become an indispensable part of many intelligent devices surrounding us, as audio is one of the most efficient ways of interacting with these devices. The accuracy and performance of KWS solutions have been the main focus of the researchers, and thanks to deep learning, substantial progress has been made in this domain. However, as the use of KWS spreads into IoT devices, energy efficiency becomes a very critical requirement besides the performance. We believe KWS solutions that would seek power optimization both in the hardware and the neural network (NN) model architecture are advantageous over many solutions in the literature where mostly the architecture side of the problem is considered. In this work, we designed an optimized KWS CNN model by considering end-to-end energy efficiency for the deployment at MAX78000, an ultra-low-power CNN accelerator. With the combined hardware and model optimization approach, we achieve 96.3\% accuracy for 12 classes while only consuming 251 uJ per inference. We compare our results with other small-footprint neural network-based KWS solutions in the literature. Additionally, we share the energy consumption of our model in power-optimized ARM Cortex-M4F to depict the effectiveness of the chosen hardware for the sake of clarity.
Comments: 5 pages, 5 figures
Subjects: Sound (cs.SD); Artificial Intelligence (cs.AI); Audio and Speech Processing (eess.AS)
Cite as: arXiv:2111.04988 [cs.SD]
  (or arXiv:2111.04988v1 [cs.SD] for this version)

Submission history

From: Osman Erman Okman [view email]
[v1] Tue, 9 Nov 2021 08:24:36 GMT (704kb,D)

Link back to: arXiv, form interface, contact.