Researchers Create First-of-its-Kind Wearable Microphone Impulse Response Data Set

Thanks to University of Illinois at Urbana-Champaign graduate student Ryan Corey and his team, scientists investigating wearable listening technology now have a novel data set to utilize.

The pioneering wearable microphone impulse response data set, being launched at the International Conference on Acoustics, Speech, and Signal Processing (ICASSP) this week, is vital to audio research for two major reasons—firstly, the data contain around 80 microphones rather than the standard two microphones demonstrating how they are heard on various parts of the body, and secondly, the data are freely available under an open-access license.

We believe hearing aids, smart headphones and all listening devices would work better if they had a lot of microphones, but most products only have two. There isn’t data out there for more than that. Even the work that has been done with more didn’t include open-access data sets.

Ryan Corey, Graduate Student, Coordinated Science Laboratory, University of Illinois at Urbana-Champaign

The data set includes over 8,000 acoustic impulse responses determined at 80 varied positions on the body. Testing of all the 80 microphones was done on five various headphone/hat styles with six different kinds of clothing. Furthermore, the sound in the recordings originated from as much as 24 different directions to replicate noisy crowds.

The group, which included former undergraduate student Naoki Tsuda and Corey’s adviser, Coordinated Science Laboratory (CSL) Professor Andrew Singer, spent many days placing all the 80 microphones over a mannequin, with Corey himself in the CSL Augmented Listening Laboratory. The team subsequently recorded the acoustic impulse responses to examine the body’s acoustics and whether or not clothes actually make a difference in the way microphones pick up noise. The data, thus collected, have been used by the researchers in the paper being presented at the ICASSP event this week; however, the team wanted the data to go farther.

We’ve been frustrated when trying to use data sets that aren’t open. Wearable arrays are important and more people should research it. Having this data out there will make it more convenient to do so.

Ryan Corey, Graduate Student, Coordinated Science Laboratory, University of Illinois at Urbana-Champaign

In the coming days, scientists can utilize the data to replicate wearable microphone arrays with varied numbers of microphones at various points on the body. A majority of the humans are already wearing numerous devices with microphones, and these data can help in exploiting that. The data can be used by engineers to design innovative products and analyze performance tradeoffs for a wide range of applications. Some of the promising applications for the data include acoustic event detection, speech recognition, and augmented reality, to name a few. Without the data set produced by the CSL group, each scientist would have to construct his or her own prototypes and test them, which is not only expensive but also time-consuming.

At ICASSP, the presentation will take place on Tuesday, May 14^th, 2019 in Brighton, United Kingdom. Corey and Singer—Fox Family Professor in Electrical and Computer Engineering—are hoping that the presentation will raise awareness of the dataset, motivate others to utilize it, and provide them a chance to receive feedback.

This is the best-attended conference for audio signal processing, so I’ll be able to introduce the data set to a lot of researchers who could potentially take advantage of it, build on it, and give us feedback for future improvement.

Ryan Corey, Graduate Student, Coordinated Science Laboratory, University of Illinois at Urbana-Champaign

The data set is available under a Creative Commons Attribution License at the University of Illinois Library’s Illinois Data Bank archives. Corey has also pended about the dataset on the laboratory’s blog.