In addition to Phonexia, researchers from the BUT Speech@FIT group at the University of Technology are behind the fame of the "Made in #brnoregion" speech technology. Do you know in which areas speech technologies help?
Voice biometrics
A voice is as unique as a fingerprint. Thus, voice biometrics can work just like dactyloscopy. Phonexia in Brno is one of the top 3 companies in the world in the development of voice biometrics. Conventional technologies can recognize a person by a specific, predefined phrase. However, this is easily exploited fraudsters can record the phrase and drop it from the recording when verifying identity.
Phonexia's technology goes much further. It doesn't care at all what language a person speaks or what they say or how they say it. Just three seconds of speech and it can identify the speaker with more than 96% certainty. And (among other things) the bank account remains safe. The company supplies its product to many countries around the world to both the government sector (police and military) and the private sector (the aforementioned banks, operators and call centres). In the government sector, they mainly help to investigate crimes and to avert various security risks.
Automatic speech recognition
Transcribing the spoken word into text used to be done manually by so-called subtitlers in television. Today, technology (thankfully) does the job for them. So today's subtitler just corrects inaccuracies or – if necessary – adds context. But this technology is also used elsewhere than in television. Transcribing the spoken word into text helps, for example, court reporters, journalists who record (and need to transcribe) hour-long interviews, analysts in call centres and, last but not least, for internal purposes in companies - for example, to take minutes of meetings.
Speech synthesis
Speech synthesis is the previous function turned on its head – it generates human speech from text. We often encounter this technology in applications such as voice assistant (voicebot). "Tin-mouth" is quite common on customer service lines, where the robot first sorts customers by request. If it's a routine matter, it can advise itself. Otherwise, it transfers the caller to a colleague in the flesh.
Spoofing & Antispoofing
With the development of artificial intelligence, there are not only more "good" tools that can save people a lot of work, but also many dark ones, such as spoofing. Imitating someone else's voice – or even making a fake video with, say, a famous person's face - will become more and more common. And it will become harder and harder to tell such videos from the real thing. That's why Phonexia is also working on so-called anti-spoofing. It has just recently received a grant for research in this area and, together with FIT BUT, will be working on technology that will be able to recognise fake voices.
Automatic speech translation
Imagine you're abroad, you get lost and need advice. Even if none of the locals (including you) speak English, you won't be lost for long. Thanks to automatic speech translation, you carry the best interpreters in your pocket (or on your mobile phone) who will translate your questions and (their) answers in high quality in real time.
Semantic search & research and sentiment analysis
Speech technologies can already analyze public sentiment, for example from social media. Likewise, they can serve companies to sense the sentiment (and disaffection) of their customers. For example, supervisors in call centres who evaluate call quality receive reports not only on the length of calls, but also on the emotions that prevailed during the call.
The Brno-based Phonexia offers semantic search as one of its functions. If you enter a keyword – say, for example, a car – the technology will find all words in the text with a similar meaning, so for example truck, passenger car or specific brands of cars. Semantic search makes the resulting interpretation of context much more accurate.
Analytical function
Speech technology offers a wide range of applications in the field of analytics – one of which is the aforementioned call quality in call centres. But speech technology also functions as a support tool for psychologists. How? They can analyse therapy sessions and give therapists very valuable feedback. After a therapy session, they can see how much space they gave their clients, who interrupted whom and, above all: if they both use the same vocabulary. This is absolutely crucial for mutual understanding and comprehension.
Voice control
Turn down the music, turn on the TV, turn on the lights, or play a favorite song? If your home is smart, your wish is granted – and you can stay on the couch. Even phones, computers and many cars are now automatically equipped with voice control. You can't stop progress. Remember, voice assistants are a good servant but a bad master. When you're happy to use them from the couch, give yourself five push-ups or ten squats before each wish just in case.
