“I want to know when event X occurs in this data!” This question is at the heart of problems faced by knowledge workers in almost every industry every day.

Events are defined by patterns or characteristics in data, and their detection is a non-trivial task. Typically the event cannot be measured directly or there are complicating factors (“noise”) that make a simple detector unfeasible. In some cases, robust, very high-reliability algorithms are needed to detect the occurrence or severity of a particular event.

Crash detection algorithms for Supplemental Inflatable Restraints (SIRs, commonly referred to as airbags) are a prime example. The algorithm does not have to analyze incoming accelerometer signals for many different events -- it must only analyze potential crash signals.

The difficulty lies in the mission-critical nature of the application. The algorithm deployed in any vehicle must have as close to 100% correct detection and classification as possible -- if it fails, lives may be lost!

Quantum Signal has experience in working with SIR algorithms and, more generally, robust detection of specific events in a continuous or discrete data, online or offline. We can work with you to build a detector for events in audio, video, recordings, text, or any other modality you wish to analyze!