Apple’s latest patent applications-like “Learning Based Distance Estimation” – focus on analyzing ambient sounds. Apple gadgets will learn how to filter useful audio information and localize its sources. It is claimed that with the advent of these technologies, smart devices will be able to interact with their owner much more accurately.
For example, the device will be able to understand from what distance it is being accessed. Depending on this, it will be able to adjust the volume. If there are several smart gadgets in the room, they will determine which one is better to answer your ” Hey, Siri!”.
It is quite easy to determine the distance to the sound source using triangulation, but this requires at least two spaced microphones. In the methods that Apple patents, a single microphone is used, but a neural network and an artificial intelligence system are taken up, supplemented by self-learning algorithms.
“The deep learning system will be able to estimate the distance to the speech source in each time frame based on the speech signals received by the compact microphone,” the patent application says.
The same technologies will allow you to recognize a person by voice and filter out noise from a useful signal. However, the Apple HomePod is already quite good at recognizing such commands while playing music.