 Hi, my name is Eva Stein, Product Manager for Vitec Mest Technique. We're here at Embedded World 2020 and I'd like to show you our solution for voice recognition and in addition to that beamforming together with our partner Voice Interconnect. So what you see right here is an application where we show how on an iMix 6 microprocessor platform a voice recognition system from Voice Interconnect is running. This is a demo application of a paper cutting machine and you can bring in voice commands and the system will react according to your voice commands. You have two microphones which can do echo cancellation to get better results of the recognition and in addition we have a system also from Voice Interconnect developed with beamforming which will listen to only noises from a certain direction either as a fixed position or depending on which noise is the highest pitch and then from there on it makes a decision where the beam is going. But more in detail, Mr. Martin Wenzel from Voice Interconnect can explain. Yes, hello, my name is Martin Wenzel, I'm co-founder of Voice Interconnect and I will show you our voice control solution here. We are a cooperation partner of Vitec and we are working together since two years now and it was quite easy for us to integrate our voice control solution onto the MIRA board. It's an iMix 6 based ARM board and our voice control runs like a service and recognizes commands and some voice prompts from customers. We have created a system which brings demo with it. So if you start the system you see already demo, you can try the voice control and you will see how good it is running. The big advantage of this system is that we run completely offline and even more we have 30 languages available so we have not only the major European languages also some Asian languages or different American languages are available. In addition to this solution there is a tool in which you can define the voice dialogue. So this is quite easily makeable. You can integrate your intents, slots and parameters. This is a scheme which is quite used in the industry right now to define a speech dialogue and the most important thing you do not have to program something in a high level language so you can just give examples, mark intents, slots and parameters and then after that you can compile a statistic language model and semantic model and you can download these models to the embedded board and then you are completely offline. And as an acoustic front end we have a beamforming here which Eve just showed you. With that beamforming you can steer an acoustic beam in a certain direction and you just listen into that direction. All the other noises outside of the beam are suppressed and this leads to far more robust functionality even under noisy conditions and also this solution is running completely offline onto the embedded system. So to sum up the advantage from this system is that you can very fast and easily create a prototype and try the solution for your product. So you've got a microphone array going on here. This is the array. The array is here, how many microphones go through and you're doing this with an IMX6. Does that provide the full microphone array like those new 20 smart speakers? It's similar to that what is integrated into smart speakers. We have compared it to other solutions and we have a superior quality I have to say that that the latest algorithms are integrated, they are fully adaptive and the software is running on the IMX6 and we have also the ability to create several beams not only just one beam so with that processor we can handle up to eight beams in parallel so that we have some advantages when you have different speakers like in a conference room where you have to record different speakers and you need different beams and you want to do a voice control or a dictation solution with that. How does it work with all the different beams? Does it automatically figure out where the speaker is and even it's far away? This leads to the question of acoustic source localization. This is a software module which is available for that solution but it has advantages and disadvantages. The acoustic source localization you typically hear to the loudest noise inside the room and if you have certain speakers in a conference solution then you want to record all the speakers in parallel and do a conference afterwards or do a dictation solution afterwards. You talk about the IMX6 platform, is that the latest newest one for voice, is it? But you do everything on this one and you say that your quality is the highest? The solution we have runs on each ARM V7 platform and above so we are not fixed to the IMX6, we can also go to the newer platforms Phytec has to offer. And here that Phytec news, have you been doing stuff with voice a lot? Well, it's the next generation of interacting with our systems. So the advantage of using for example an IMX6 or any other embedded platform is that you have an alternative way to interact with the system in addition to typical touch display interfacing, HMI kind of interfacing. So for example if you don't have the ability to be touching a system because you're for example a doctor and you're operating and you want to have your equipment react to what the situation is requiring right now then you can just talk to the system and the system will react, right? And so these are kind of new ways to interact with embedded systems and running this on an IMX6 or a ARM has the advantage that you still have the typical embedded functionality like displays, like IO connections, like interfacing with touch and seeing what is happening and in addition being able to talk to the system and manage it from this way, right? Voice Interconnect works more than 15 years in the field of signal processing so in the broadest sense, so we do a lot in the field of communication like intercom solutions and also since the beginning we have our own voice control solutions and our technology is not a hipster technology so it's field proven, we have integrated into several solutions already on the market and even more the acoustic language models we use they are really industrial, great quality we have 30 language models available and all of them have the same quality, this is something which I have to point out So that's why we brought up this partnership together with Voice Interconnect because the company Voice Interconnect is from the strategy very compatible to what we are doing, especially ruggedized and solid solutions for industrial use so through our partnership we can offer a full on solution we take care of all the hardware part and the basic operating system support and Voice Interconnect on the other side can take over the tasks when it comes to... There are a lot of kind of services necessary which we are providing to integrate voice control successfully into a product It starts with the acoustic engineering so typically you have to integrate the microphone into your product, into your housing and we can give our customers support with some consultation services to how to integrate the microphone into the housing and in addition we can adopt the solution to specific needs so we can integrate some software functionalities if necessary and there is a big trend right now, a lot of voice talk in the industry and you've been doing it for 15 years so does that mean you invented a lot of it or are you working with some famous brands out there or doing some stuff based on your work? In the beginning we were developing our own recognizers in German and English and afterwards we started selling these developed algorithms and we were working for the automotive industry where the first applications with voice control were seen there like a hands-free car kit and we have done a lot of things developing such hands-free car kits and afterwards we started the white goods industry so household appliances like a voice control coffee machine from Jura this was one of the first voice control products in 2010 Do you think a lot of the world would be voice controlled in the future? Is it really going to happen? Voice control has several advantages in controlling a product so you have hands-free, you have eyes-free and also you are more efficient as you can say several commands in one row so you do not have to go on a touch screen into several sub menus and do settings there you can say everything what you want to control in one row and this is an advantage over a standard graphical user interface