Latest revision as of 11:24, 14 May 2024

Status: Available

• Type: Semester Thesis
• Professor: Prof. Dr. L. Benini
• Supervisors:
  • Philip Wiese (IIS): wiesep@iis.ee.ethz.ch
  • Victor Kartsch Morinigo (IIS): victor.kartsch@iis.ee.ethz.ch

Introduction

Sound based vehicle classification and counting than could be combined with air pollution monitoring to perform multi-modal sensing.

Smart sensors, which are sensors equipped with advanced computation capabilities, present a unique opportunity for innovation in multiple critical domains such as environmental sensing to reduce global greenhouse gas emissions or infrastructure monitoring. Our infrastructure is responsible for almost 80% of all global greenhouse gas emissions. Therefore, it is very important to better understand the individual contributors and to be able to build precise models to plan future infrastructure.

A near-sensor processing of the gathered data that directly extracts meaning from the signal can lead to huge improvements in terms of energy efficiency and increased privacy. The reduced energy consumption comes from the fact that more specialized hardware such as the PULP platform including the GAP9 can be used, and, additionally, that the information density of the transmitted data is higher, and thus less data must be transmitted via the often very power-intensive wireless link.

Our long-term goal is to develop a low-power sensing node with very low power consumption. These nodes should incorporate multiple sensors to enable multi-modal sensing. In other words, combine the data of the different sensors to extract additional data. The idea is to use the very efficient GAP9 multicore processor and deploy multi-modal neural networks to perform feature extraction, prediction, anomaly detection, and even on-device learning to adaptively fine-tune the model based on the measured data.

Vehicle counting is a critical tool for urban planning and environmental monitoring, providing essential data on traffic flow and its associated impacts. This data is vital for understanding the relationship between vehicle traffic and air pollutants, which helps assess environmental health risks. Accurate vehicle counts are crucial for managing traffic in constrained spaces such as tunnels, where excessive occupancy can pose safety risks. Vehicle counting also aids in planning road infrastructure improvements and in implementing smart city initiatives aimed at reducing traffic congestion. The use of microphones for vehicle detection offers several benefits over traditional methods, such as cameras. Microphones consume less power, making them more energy-efficient and cost-effective for continuous monitoring. Furthermore, because they rely on sound rather than visual recordings, they preserve anonymity and are less intrusive, addressing privacy concerns more effectively than camera-based systems. The challenge of accurately identifying and counting vehicles based solely on sound involves using one or more microphones to detect the type and number of vehicles passing by. This approach can be extended with a multi-microphone setup to determine the vehicles' driving direction and speed. Students will work on implementing such a system for raw data acquisition and deploying a neural network trained on a publicly available dataset and fine-tuned on a dataset acquired with the device. Besides counting, the system also aims to determine driving direction and speed. On top of this, further research can be done by integrating a gas sensor to measure pollution levels, allowing for an analysis of the correlation between traffic density and air pollution, and providing further insights into environmental impacts.

Project

• Data Acquisition: Development of a platform for the acquisition of multiple signal sources and creation of a dataset.
• Algorithm and Network Architecture: Benchmarking of different algorithms and neural networks based on the current literature.
• Implementation: Implementation of algorithm on GAP9.

Prerequisites

• Good understanding of deep learning concepts.
• Familiarity with Python programming.
• Familiarity with the PyTorch deep learning framework.
• PCB Design and embedded C programming.

References