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

Download:

Current browse context:

eess.SP

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

Electrical Engineering and Systems Science > Signal Processing

Title: Resource Constrained Neural Networks for 5G Direction-of-Arrival Estimation in Micro-controllers

Abstract: With the introduction of shared spectrum sensing and beam-forming based multi-antenna transceivers, 5G networks demand spectrum sensing to identify opportunities in time, frequency, and spatial domains. Narrow beam-forming makes it difficult to have spatial sensing (direction-of-arrival, DoA, estimation) in a centralized manner, and with the evolution of paradigms such as artificial intelligence of Things (AIOT), ultra-reliable low latency communication (URLLC) services and distributed networks, intelligence for edge devices (Edge-AI) is highly desirable. It helps to reduce the data-communication overhead compared to cloud-AI-centric networks and is more secure and free from scalability limitations. However, achieving desired functional accuracy is a challenge on edge devices such as microcontroller units (MCU) due to area, memory, and power constraints. In this work, we propose low complexity neural network-based algorithm for accurate DoA estimation and its efficient mapping on the off-the-self MCUs. An ad-hoc graphical-user interface (GUI) is developed to configure the STM32 NUCLEO-H743ZI2 MCU with the proposed algorithm and to validate its functionality. The performance of the proposed algorithm is analyzed for different signal-to-noise ratios (SNR), word-length, the number of antennas, and DoA resolution. In-depth experimental results show that it outperforms the conventional statistical spatial sensing approach.
Subjects: Signal Processing (eess.SP)
Cite as: arXiv:2107.11070 [eess.SP]
  (or arXiv:2107.11070v1 [eess.SP] for this version)

Submission history

From: Shivam Chandhok [view email]
[v1] Fri, 23 Jul 2021 08:25:35 GMT (2295kb,D)

Link back to: arXiv, form interface, contact.