Hardware Accelerator and Neural Network Co-Optimization for Ultra-Low-Power Audio Processing Devices

TL;DR

HANNAH is an automated co-optimization framework that designs neural networks and hardware accelerators together, achieving ultra-low-power audio processing with high accuracy on resource-constrained edge devices.

Contribution

The paper introduces HANNAH, a novel framework combining neural network and hardware co-design using evolutionary algorithms and analytical models for ultra-low-power audio applications.

Findings

HANNAH finds neural network and hardware configurations with minimized power consumption.

HANNAH achieves higher accuracy in audio classification tasks compared to existing methods.

The framework effectively balances power, performance, and accuracy for edge devices.

Abstract

The increasing spread of artificial neural networks does not stop at ultralow-power edge devices. However, these very often have high computational demand and require specialized hardware accelerators to ensure the design meets power and performance constraints. The manual optimization of neural networks along with the corresponding hardware accelerators can be very challenging. This paper presents HANNAH (Hardware Accelerator and Neural Network seArcH), a framework for automated and combined hardware/software co-design of deep neural networks and hardware accelerators for resource and power-constrained edge devices. The optimization approach uses an evolution-based search algorithm, a neural network template technique, and analytical KPI models for the configurable UltraTrail hardware accelerator template to find an optimized neural network and accelerator configuration. We demonstrate that HANNAH can find suitable neural networks with minimized power consumption and high accuracy for different audio classification tasks such as single-class wake word detection, multi-class keyword detection, and voice activity detection, which are superior to the related work.