Well, to the computer systems, do you really think that the sounds of different objects do not matter?A new AI is being taught how to listen and differentiate sounds of different types. In the sound proof hanger of RAF airbase, Chris Mitchell CEO and founder of Audio Analytic and his colleagues are breaking innumerable glasses to teach computers a lesson.
What lesson is it? Welcome to the latest Artificial Intelligence organization, a Cambridge-based startup company which is focused on training machines to learn and recognize sounds of breaking glass. The team gathered thousands of window panes and doors of different sizes and shapes and started to smash them.
The training
The company is recording the different shattering sound of each type of glass. Sometimes that are swinging the sledgehammers or the garden spades, and sometimes they are throwing bricks. The team also worked with vets to track the sound of barking dogs of as many as possible of different breeds. The company is teaching to pick up sounds that are essential for humans, like the sound of smoke alarms, babies and also barking dogs. The fundamental objective is to recognize sounds without confusing one sound with other. You just cannot deny that AIs are gradually getting better than humans at identifying different objects, especially the faces.
How is Audio Analytic unique?
If you consider the different AIs available in the market, like Apple Siri, Google Home and Amazon Alexa, the companies didn’t give proper attention to the sound recognition feature. Most of these devices take every day noises for most of the machines. The company plans to change that and this is why they are working in this new field of AI. The Audio Analytic is that part of new wave enterprises that spot the different patterns in sounds. Also, there is another company called Uberchord, it is based in Berlin and is developing the AI that will help people in learning to play the guitar. It will listen to you and tell you when you are placing the wrong finger in wrong chords.
Few other companies
Another company, the Cambridge Consultants, taught the AI to recognize different genres of piano music like the ragtime or the baroque. According to Audio Analytic, if the smart home devices can differentiate between a dropped wine glass and a smashed window; it can alert the smart homes when an intruder breaks in. The company has bigger ambitions; they are focused on creating the taxonomy of all sounds and this is a major project. At present, the company’s software can identify breaking windows, the crying babies and also the smoke alarms. They also added barking dogs in their latest report.
Also Read: Baidu Robot Xiaodu Challenges Humans In China
The additional functionalities
The Audio Analytic is also working on the anomaly detector. It will pick up sounds that are not an ordinary one like the clatter of something falling, the hiss of water pipe. Eventually, the company is going to add car alarms. For the US market, they might also add the gunshots. The ability to recognize different sounds matter. According to the company, “It is fast becoming vital for smart home technology to detect and interpret a broad range of ambient sounds, so people can respond easily and quickly to what’s going on in their homes when they’re not there. You could be alerted if your smoke alarm was going off while you’re in a meeting at work, and not just when you’re on your sofa.”
Audio Analytic is working on
Apart from sending the alert message to your mobile phones, such systems can take actions by themselves. For instance, a smashed window can turn on the lights; a baby’s cry can start the lullaby from a nearby speaker. The tough part is ensuring whether the AI can identify what it is hearing. Because false alarms can create a nuisance for the buyers. To get enough recordings the company worked with the parents’ groups in Cambridge. Then they came up with their lexicon to describe the different types of cry. To teach smoke alarms, Audio Analytic bought as many different models as they could online. At first, the company faced trouble in reading the beeps of ringing phones, the alarm clocks and the oven timers. This is why they trained it to focus on a single alarm pitch within the beeps.
The Audio Analytic is working on something really advanced tech. Let us wait and watch how it works out for the consumers and we will update you more about it.
