#IMAGINE Lab, Center for Spatial Planning Analytics & Visualization & IPDL (Interactive Product Design Lab)
#project grant from 2017 IPaT Smart & Connected Communities Data Pilot Grants
#Category: #Smart Cities, #IoT, #Wireless Sensor Network (WSN), #Wireless RF, #Acoustic Activity Detection
Adviser: Matthew Swarts
Date: Fall 2017 - Present
Weatherbug is about designing an environmental sensing module/device which is constantly monitoring environmental conditions in the area to collect, store, transmit and process data. Collected data are temperature, air pressure, humidity, wind speed, light and sound. This proposed sensor device is solar-powered. It collects data in real-time and periodically transmits the data using a long range, low power wireless platform called LoRa. Twenty of these modules will be spatially distributed in a network across campus. Such open platform sensor device not only assists environmental conditions monitoring but also, allows more easily exploration and test of novel algorithms for event detections at the urban scale.
We have started developing this work-in progress project, by focusing on three areas first: collection of sound data, transmit data with LoRa, and designing a specific circuit to measure wind speed and temperature.
Overlapping sound event detection|
Recognizing and locating individuals, vehicles and other sound events has a large applicability. Simplified situations for sound event detection including isolated sounds is not realistic for our desired applications. Considering the complexity of the audio surrounding us in everyday life, the detection of overlapping sounds continues to be a challenging problem. To detect sound events around us, we are using Fast Fourier Transform. Diagram x demonstrates the workflow.
Fast Fourier Transform|
Complex signals made from the sum of sine waves are all around us. The Fourier transform gives us insight into what sine wave frequencies make up a signal. In this project, we are applying knowledge of the frequency domain from the Fourier transform in audio processing and detecting specific tones and frequencies to detect when and for how long each target event is happening. According to Nyquist’s sampling theorem, we are using two parameters; sample rate and fft size.