During bracelet usage, the microphone records all sounds that exceed the predefined threshold, so it is necessary to filter those sounds which present potential threat to a user. For example threat sounds can be:
- Fire alarm
- Care horn
- Police, ambulance and firefighter siren
- Fire and smoke alarm (within building)
- Dog growling and barking.
In the Matlab software package, we performed a spectral analysis of the sound that produces smoke alarm. Furthermore, we plan to do an analysis of all signals that are of interest to bracelet users ie, signals that can pose a danger / threat to the user. When we collect and analyze all results, we will decide which type and how we will implement the digital filter.
As an additional bracelet functionality we are considering to perform classification of the signal, which would enable the bracelet to "recognize" the sound and initiate different vibration intensity/duration for different signals. For example, if a bracelet detects a police siren, it can vibrate constantly and thus increase certainty that user will not miss the warning.