Cisco has expressed its intent to acquire BabbleLabs, aiming to solve a problem that professionals often face — unwanted background noise, constant interruptions and disturbances while they are on a video conference. The acquisition will integrate intelligence into the Cisco Collaboration platform.
BabbleLabs released this tech back in 2017 and uses supervised machine learning models that were trained on things like dogs barking, keyboard clicks, sirens and so on. With this, the company can distinguish human voice from the background noise and then remove that background noise in real-time with extremely low latency and great performance. This feature is an output of a combination of machine learning and clever audio techniques that can be accessed via its APIs.
The Strategic Advantage For Cisco In This Acquisition
The BabbleLabs acquisition furthers Cisco’s investments in AI innovation and talent over the last three years which has focused primarily on making Collaboration more effective and secure.
As per a Cisco global survey centred on the future of work, 98% of workers reported they encounter frustration from disturbances during video meetings when working from home. In fact, two of the leading five frustrations told are concerning backdrop noise— either from other participants or their individual side during a team call.
Working from home will become more prominent going forward, and Cisco clearly wants to provide the best video meeting experience from anywhere across all devices through the Webex application.
With the addition of BabbleLabs, Cisco will deliver native noise removal ability to its whole Collaboration portfolio. Originally, Cisco will focus on combining BabbleLabs to produce a state-of-the-art audio experience to Webex Meetings users across all devices. Webex was acquired by Cisco in 2007 and provided products such as Webex Meetings, Webex Teams, Training Center and more. Many people in the IT sector speak highly of WebEx for its unique experience.
The value proposition of BabbleLabs is that by adding this kind of software processing to speech-based systems whether it’s telephones or conference calls, Cisco can reduce the cost, improve the recognition rate, while improving clarity and productivity for the participants. The technology also preserves user Privacy and Cloud Security by processing noise removal 100 per cent at the source where noise occurs (the client-side), which aligns with Cisco’s policy of security by design.
Javed Khan said, “We are happy to announce the intent to acquire privately-held BabbleLabs, which is at the forefront of utilising advanced AI techniques to detect human speech from any unwanted noise. That means vacuums, sirens and other diverting noises are silenced, and your voice comes through clearly. With the addition of BabbleLabs, we will bring this native noise removal ability to our Webex Collaboration portfolio.
Using AI For Noise Elimination
Using advanced deep learning, speech science and digital signal processing technology, the software enhances speech and removes disruptive sounds such as keyboard typing, reverberation, kids yelling, HVAC systems, dogs barking, kitchen appliances and more. This effectively eliminates noise at the consumer level and for businesses that rely deeply on the clarity of communication as enterprises work to communicate and collaborate internally with their customers, suppliers and partners.
That’s high-value communication where clarity is of the essence for business success. By using deep learning in such a sophisticated way, the company aims to remove much more noise than ever before. BabbleLabs is at the cutting-edge of dealing with a noisy, reverberant environment and making speech work for phone calls, video recording, and user interfaces in all kinds of devices and platforms. It’s not just for a specific platform, or specific use case.
How Does This Compare To Rival Apps?
BabbleLabs is taking a different approach to speech processing. We are all familiar with the rapid emergence of speech in recent years with Siri and Alexa and Google cloud voice and other speech services that are largely processed in the cloud. They also are focused on taking continuous speech streams and delivering them to applications, then sorting out some specific tasks in mind. Instead, BabbleLabs’ technology only looks for specific tasks that the application cares about and as a result disambiguates right from the beginning, allowing the company to get to impressive recognition rates at considerably less compute and memory than the cloud-based systems.
In fact, the use of deep learning to speech quality has caused a huge change in what advanced software can do for a noisy speech. BabbleLabs recently researched, performed and measured the state-of-the-art noise reduction algorithms over the last 40 years utilising a uniform metric for quality (PESQ: ITU P.862) and a consistent speech data set (Interspeech 2020 DNS challenge). The company says it found that in the past two years, its deep learning algorithms have surpassed all earlier methods. BabbleLabs moves beyond the current noise suppression solutions by separating speech from background noise; eliminating background noise in real-time; and improving the voice to promote communication, irrespective of the language.
On the other hand, one can also make an argument that WebEx is essentially playing catch up with rival apps such as Zoom and Microsoft Teams. The current version of Zoom already has background noise suppression, which can help remove distracting noises already built-in. The same goes for MS, which already uses AI to remove background noise from calls automatically since March when the company reported daily active users going above 40% in just a week. Overall this is a good move for Cisco and can prove to be a helpful innovation for WebEx amid pandemic.
That Webex platform is important to Cisco and its customers and integration of BabbleLabs technology into the Webex family could have a rapid and profound impact. Moreover, the span of Cisco’s AI innovation across speech enhancement, speech recognition and speech analytics meshes closely with the key innovations already underway within the Cisco Collaboration teams. The acquisition is anticipated to finish in the first quarter of Cisco’s FY21, subjected to conventional closing provisions and needed regulatory approvals.
If you loved this story, do join our Telegram Community.
Also, you can write for us and be one of the 500+ experts who have contributed stories at AIM. Share your nominations here.
Vishal Chawla is a senior tech journalist at Analytics India Magazine and writes about AI, data analytics, cybersecurity, cloud computing, and blockchain. Vishal also hosts AIM's video podcast called Simulated Reality- featuring tech leaders, AI experts, and innovative startups of India.