March 17, 2021
By Pavel Jiřík in Blog
Today is a very special day for Phonexia. We are celebrating our 15th birthday! It is incredible how time flies. On March 17, 5,479 days ago, a group of excellent researchers decided to pursue the intriguing mysteries of voice biometrics and speech recognition. They founded Phonexia, and the world of voice recognition solutions has never been the same since.
15 Years of Innovation
2006 was quite a year to remember. Facebook and Twitter introduced social networks to the general public and transformed the way people interacted with their families and friends. The Project Purple 2 was finished just in time to be unveiled as the first iPhone only nine days into 2007, revolutionizing the concept of mobile phones.
Yes, 2006 felt magical. And it was the perfect year for Phonexia to be born.
Starting with only a handful of employees, Phonexia has grown into a market-leading vendor in the voice recognition industry and has collected many prestigious awards along the way. Continually ranking among the top six winning teams in the speaker recognition competitions organized by the National Institute of Standards and Technology (NIST), Phonexia has been transforming the latest scientific breakthroughs into the everyday reality of cutting-edge commercial products.
Fast forward to 2018, Phonexia was already experimenting with advanced deep neural networks and became the first vendor in the world to release a voice biometrics engine for automatic speaker recognition that was entirely powered by deep neural networks—Deep Embeddings™. The solution offered unmatched performance in terms of accuracy and speed.
Then, in 2019, our research team improved the accuracy of Deep Embeddings™ even further and broke through the mythical barrier of 99% speaker recognition accuracy on the NIST SRE dataset, achieving a fantastic 0.96% Equal Error Rate.
Last year, our researchers massively improved the accuracy of our speaker recognition engine, offering over 92% verification accuracy from out of the box on just three seconds of speech. And to finish the year off, the Phonexia research team won first place in the dialect identification task of the Oriental Language Recognition 2020 Challenge.
Yes, the last 15 years of Phonexia have been an extremely fascinating ride!
The World's Top Voice Biometrics Expert Niko Brummer Joins Phonexia
What better present could we wish for today’s 15 years of Phonexia than having Niko Brummer, the world’s top voice biometrics expert, join Phonexia!
Niko Brummer has been immersed in voice biometrics research for more than two decades, working on many interesting voice recognition projects for companies such as Agnitio, Nuance, and Omilia, finding answers to many puzzling questions around the human voice. He is one of the best voice biometrics researchers in the world, and we are very happy to welcome him aboard today!
What sparked your interest in voice biometrics?
“In the late 90s, a banking call center approached my employer, asking for a text-independent speaker verification solution. Only text-dependent solutions were commercially available then. We had no such technology ready to offer, but I was given a research paper and told to ‘build one of those’. I did so, and we eventually installed a successful demo, running live at the call center. Unfortunately, the Y2K bug diverted the focus of the bank and the final installation was cancelled. Despite this setback, I liked this problem and its ever-evolving solutions so much that it became the main focus of my career from then on.”
What amazes you the most about voice biometrics?
“That it works. That it works better than humans in many situations. That it still continues to improve in accuracy despite decades of research.”
What did your first voice biometrics solution help to solve?
“Referring to my first answer, text-independent speaker verification is when the to-be-verified human subject can say anything, as opposed to text-dependent modality, where a prescribed phrase must be said. In text-independent mode, the speech samples can be collected during a natural conversation between the to-be-verified caller and the call center agent. The call center management believed this would provide an easier and more acceptable experience for their clients.”
What do you think is the biggest voice biometrics challenge lurking ahead?
“To equal and perhaps exceed human performance in adverse conditions, where speech samples may be short, noisy, and contaminated with speech from one or more other speakers.”
How do you think voice biometrics will evolve in the upcoming future?
“It will become more accurate, especially in challenging conditions. And it will become more widely available, until most machines that are equipped with speech technology will include speaker recognition by default."
What made you decide to join Phonexia?
“The good voice recognition technology and a friendly culture.”
What do you envision helping Phonexia achieve in the upcoming years?
“I hope I can make a big contribution to help Phonexia to meet the above-mentioned challenges and to make progress towards the future.”
What would your ideal version of the world powered by voice biometrics look like?
“I'd like to see that speech-enabled man-machine interfaces become much more common, and all of them should be able to recognize speakers.”
Niko perfectly complements our vision of a voice-first world, and we are looking forward to pushing the boundaries of voice recognition capabilities together, until the vision becomes a reality.
The Future of Phonexia (and Voice)
Phonexia has always pursued its ambitious goal of making the world a better, safer, and more intuitive place to live. This has been true for the past 15 years and will continue to be increasingly so in the years to come.
We are dedicated to solving everyday challenges through voice and letting millions of people enjoy a voice-first universe.
A voice-first experience that is powered by extremely fast and accurate voice verification, precise speech transcription, and smart conversational AI that interacts with the person at any moment through highly personalized responses is the goal that fuels our Phonexia spirit.
As Phonexia CEO Michal Hrabi says about a voice-first future:
“As a big fan of sci-fi, I always like to think about how the human voice is going to be used in the future. For me, a voice-first future is about the ease of interaction between people and technology. It is about personalized voice interactions adjusted uniquely to the person's needs. For example, if I am in a meeting room with other people and look at the projector screen and say ‘What is my agenda for the next week?’, my personal agenda will instantly show up and not the agenda of someone else in the room.”
The future holds so much potential for voice recognition technologies, and we at Phonexia are so ready to transform today’s sci-fi into the reality of tomorrow!