With the pandemic having changed work life as we know it, many of us have used programs like Zoom or Slack to call our colleagues and discuss projects. But we’ve also all faced the noisy challenges that come along with these calls — a colleague’s dog barking in the background, your neighbor loudly mowing their lawn, or even just someone on your team having a particularly noisy keyboard.
Some calling programs have introduced noise-canceling AI as an option for users to reduce background noise in their virtual calls and make these calls more efficient and productive. Noise suppression has been an optional setting in many calling apps for a number of years but very rarely is it 100% effective.
How Does Noise Cancelling AI Work?
Essentially, the noise-canceling AI has to learn the difference between the sound of a human voice and any sort of background noise that it may encounter. This distinction can be difficult for a machine to recognize under normal circumstances, as the human voice shares a similar frequency with many other natural sounds.
When AI knows what sounds it needs to suppress, it can then employ active noise-canceling to eliminate that background noise.
Active noise-canceling happens when a machine detects the noise it needs to cancel, then generates its own noise that is out of phase with the background noise. The two sound waves effectively cancel each other out, both becoming inaudible.
A graphical depiction of active noise cancellation. Image/Wikimedia Commons
Noise suppression software that employs active noise-canceling is often employed in headsets at construction sites and other jobs that require working in a very loud area. Noise-canceling AI can employ this tactic to make online calls clearer.
Leaders in Noise Suppression Software
Noise-canceling AI has come a long way in recent years, especially amid the pandemic. In late 2020, Microsoft introduced noise suppression software in the Microsoft Teams work platform by developing AI to detect and cancel out background noise.
To do this, Microsoft created a dataset with nearly 1,000 hours of noise data, differentiating between clean speech and background noise. They recorded speech in over 10 languages and recorded over 150 types of background noise to ensure the technology works in nearly any situation. They fed this data to their noise-canceling AI to teach it to differentiate between the speech and background noise.
As a result, the AI is able to suppress background noise while still allowing clean, clear speech to come through to the user.
While Microsoft’s solution is exclusive to its own conferencing platform, a newer AI company, Krisp, was founded in 2017 and aims to provide its noise suppression software to anyone. It works on over 800 existing platforms like Zoom, Webex, and Skype, rather than requiring one specific infrastructure to function.
Krisp provides plans for both individual users and companies and even provides its individual users with noise suppression on both ends. This means that not only will background noise on the user’s side be suppressed, but they won’t hear the noise from the other person’s end, even if the other person doesn’t have the program installed. Krisp’s AI has learned to recognize and cancel out background noise on the fly, almost instantly as the sound is transmitted.
Nvidia is currently working to use deep learning to improve the way technology handles noise suppression. Its software RTX voice has an AI-driven two-end noise suppression system that is similar to Krisp, although it can only run on Windows 10 and requires an Nvidia RTX graphics card. With both large and small software developers stepping into noise suppression software, the technology has evolved rapidly.
Noise Cancelling AI in Restaurants
As businesses shift from human operations to technological solutions like self-ordering kiosks, it’s essential for any voice recognition software to be able to effectively filter out the background noise and recognize the customers’ speech without error.
If a customer tells a machine that they want a double cheeseburger and ends up with a chicken sandwich because the background noise muddled the speech recognition, they aren’t going to be happy.
The same principle applies to conversational ordering kiosks, which speak directly to customers and allow them to order completely verbally, without having to touch the kiosk at all.
These kiosks are designed to imitate human speech patterns and to ensure customer satisfaction, they will need to be able to hear the customer completely and accurately at all times. This becomes possible by implementing noise-canceling AI in the kiosks.
The benefits of conversational AI mostly emanate from its advanced speech recognition and ability to imitate speech fluctuations and thus sound more natural.
Customers may become frustrated or even uncomfortable dealing with a monotone robot that can only recognize certain keywords, but implementing conversational AI allows for more dynamic, fluid, and pleasant experiences for customers.
Conversational AI kiosks can even be trained to recognize faces, allowing them to identify characteristics in a customer that may impact what they want, or even recognize a repeat customer and remember their order history, enabling the AI to make personalized recommendations and upsells.
Further, with labor shortages among food service industry workers becoming more prevalent, conversational AI restaurants that implement noise suppression software could entirely eliminate the need for front-of-house staff.
The concept of restaurants and other businesses implementing AI POS may sound like a thing of the future, but it exists today, and it’s already helping businesses maximize their profits. Over the past few years, major food chains like McDonald’s and Panera have implemented kiosks in their restaurants, leading to increased sales and revenue. Implementing AI POS is just the next step in maximizing efficiency and customer satisfaction.
Combining noise-canceling AI and conversational AI can even pave the way for conversational AI restaurants, which would be able to take orders and answer questions from customers without any employee involvement.
The Future of Noise-Cancelling AI
These recent improvements in noise-canceling AI can vastly improve the efficiency and productivity of video calls, and there’s still plenty of space to improve this tech and expand its usefulness.
Additionally, a store could conceivably have multiple conversational ordering kiosks implemented inside, allowing for the quick servicing of multiple customers. Combining this with noise-canceling AI would prevent any miscommunications, such as one kiosk overhearing a customer’s order at an adjacent kiosk and getting the orders mixed up.
All of these technologies combined would create a customer service experience that boosts speed and ease of ordering while minimizing errors and miscommunications. More customer satisfaction means more returning customers, which means higher revenue for restaurant owners. The implementation of AI-powered kiosks also cuts down on the number of employees needed to operate the restaurant.
Effective, consistent, and customizable noise cancellation software could allow for online music lessons to become easier for instructors to manage, or allow live-streamed music performances to be clearer, with less echo and better sound quality.
In an increasingly automated world, noise suppression software will also become a larger necessity in everyday situations. Using voice commands on your phone on a noisy city street could become much easier with AI noise cancellation technology.
With the unlimited possibilities that AI provides for businesses — from dynamic noise suppression software to conversational kiosks that can greet a returning customer by name – nothing should stop business owners from taking advantage of modern technology to enhance their business’ efficiency and competitiveness.