EngineEar

IBM research had been working on acoustic recognition software using deep learning techniques and were exploring industrial use cases.

The technology samples the environment via a microphone and extracts a range of features such as the amplitude, pitch and timbre (sound texture), and stores these as a sound library. A user enriches the library by listening to the samples and labelling them, for example by indicating whether the machine was healthy and the type of failure if not. The solution could then return status by finding the closest match to those in the library, for example sending and alert of machine failure and likely cause.

Chris identified the technology as having potential for providing automated remote monitoring of plant and machinery in buildings and infrastructure. He secured the first commercial pilot globally of the solution with Interserve, an outsourcing services business who were interested in the potential for condition-based and predictive maintenance. They ran the pilot at a factory and he secured other pilots including at the plant room of a global bank's European headquarters and at one of the UK's largest waste water treatment facilities.

The solution was eventually launched as IBM Acoustic Insights.