Human hearing is very good when it comes to filtering any sound against a background of general noise. We can easily single out the voice of the interlocutor, even if there are ten other people talking in the room besides us. Computer systems like Siri do not work so well, because they simply do not know how to perfectly separate one source of sound from others. Researchers from the University of Duke (USA) A new sensor that should solve this problem is modern gadgets.
To create such a sensor, scientists had to use the so-called "metamaterials". They are called so because they can boast properties that are not found in nature. Also, the researchers used a special technique of sound processing, which is called "Compressive sensing" and is a technique for obtaining and restoring a signal, using knowledge of its previous values.
The device was disc-shaped and not equipped with any electronic or moving parts. Instead, it has a structure that resembles a honeycomb of varying depth and shape. These cells, located in the form of special patterns, allow the sensor to acquire such an incredible "rumor". As a result of distortions obtained when sound passes through these "honeycombs", sound acquires unique characteristics. And when he finally reaches the microphone in the center of the sensor, the computer algorithm, taking into account the distortions obtained, can accurately separate the different sounds from each other and even determine from which side they entered the sensor.
Scientists argue that the sensor is able to almost perfectly separate the sounds obtained simultaneously from several different directions. Three sounds from different sources were separated with an accuracy of 96.7%. Now the sensor has a diameter of about 15 centimeters, but the researchers promise to reduce its size so that it can be integrated into a wide variety of electronic devices, including smartphones and tablets.
The article is based on materials .
Comments
Post a Comment