A company called Cypher has found a way to put a new twist on an old technology in order to help various voice based AI assistants serve you better.
The technology in question is noise cancellation, and the twist is not some type of new hardware, but instead a unique software tweak.
This software tweak works by incorporating a Deep Neural Network into existing noise cancellation softwares, so that they could apply machine learning to it.
The machine learning program differs from existing noise cancellation technologies in not just accuracy, but also in how it goes about achieving that accuracy.
While traditional programs focus primarily on trying to identify the noise and subtracting it from the speech, this new program focusing on subtracting the speech from the noise.
This makes the whole process a lot easier, as the human voice has some very distinct patterns that make it easier to recognize when compared to other types of abstract noises.
This program is so effective that the company claims that it can cut out almost 99% of background noise with a delay of just 24 milliseconds in real time performance.
This is a huge improvement over other Intelligent Noise Cancellation products, who can neither achieve this level of accuracy, nor hold it for a long period of time.
The company plans to use this new method of noise cancellation, in order to improve the accuracy of speech recognition among AI softwares such as Siri, Cortana and others like them.
In doing so, they will be able to help these assistants perform a lot better than their current state, especially when it comes to various nationalities and accents.
It’s still not clear which AI will get this new advancement first, but no matter which one does, we can’t wait to see the results for ourselves.