The HumBug technical team has worked to bring a series of updates and innovations to the heart of the project. Here are some of the recent highlights.
1. Detection and species classification: Close collaboration with the data science team from Berkeley has accelerated our detection and species classification work. Not only have the team consolidated improvements in real-world detection of mosquito sound events, but - and importantly - the ability to accurately determine the likely species of mosquito has been significantly improved. This is coupled with foundational work on the data pipeline, allowing automation from audio data through data cleaning to mosquito detection and species likelihoods.
2. The MozzWear app: At the core of our data collection methodology lies the MozzWear app, a robust phone app which seamlessly records, compresses and transmits audio data to our central servers. Recent changes to the app have pushed several improvements, allowing enhanced control of timers for the recordings as well as porting the Android app to the iOS environment. We have even managed to have a prototype version of the mosquito event detection algorithm working successfully in real time on a phone, paving the way for distributed analysis and the potential for phone-specific feedback to users.