Contextual Intelligence on Resource-constraint Bluetooth LE IoT Devices
Despite the general acknowledgment by the scientific community that “Genetics load the gun but environment pulls the trigger”, there is persistent uncertainty as to the global burden of disease attributable to environmental (including life-style and climatic) factors, including healthcare costs and negative economic impact. Given that we spend 60-90% indoors, wiretapping the indoor environment and collect individual-level exposure data is a crucial aspect in keeping the workforce and our loved ones safe and healthy. Apart from gas concentrations, sensor data of interest include: • Daylight exposure: Natural light, Glare and UV • Thermal comfort: Temperature and humidity • Noise pollution: Sound Levels and Audio signature classification • Electromagnetic radiation With interminable technology advances, it is now possible to assemble multi-model IoT sensors based on solid-state electronics including low-power CMOS gas sensors and MEMS microphones on a small footprint. The aim of the internship is to explore efficient ways to extract relevant information (Sensor Fusion) in real-time on resource-constraint, Bluetooth Low Energy IoT devices. While all of the above indoor environment impacts are of interest, a key focus of this project will be on audio as the acoustic sensor (microphone) offers a medium-bandwidth spectrum well suited for embedded time-frequency analysis, potentially augmented with machine learning capabilities based on an annotated audio training set. Thus, a key goal of the internship is to implement and evaluate (1) A-weighting sound pressure level algorithm(s) and (2) explore audio classification algorithms running on embedded systems.
This project is done in collaboration with Honeywell Switzerland.
According to the skills and attitude of the student, the project will include the following tasks/activities
- Familiarize with standard A-weighting sound pressure level algorithms, and derive an evaluation plan
- Perform state-of-the-art survey on audio classification algorithms running on resource-constraint hardware, and select suitable approaches for this project
- Implement the A-weighting sound pressure level algorithms on Nordic’s Thingy52 IoT dev kits (or similar with ARM-Cortex-M processors) and evaluate the performance
- Explore and evaluate audio classification algorithms on the Thingy52 (or similar)
- Looking for Semester and Master Project Students
- Supervisors: Michele Magno
(not all need to be met by the single candidate)
- Experience using the laboratory instrumentation - signal generators, oscilloscopes, DAQ cards, Matlab etc.
- Basic knowledge of communication systems
- knowledge of microcontroller programming (C, MP-ARM, ARM-Cortex-M, Nordic BTLE)
- basic knowledge on signal processing and machine learning is a plus.
- plus is knowledge on printed circuit board (PCB) using Altium.
- 25% Theory
- 55% Implementation
- 20% Testing