with blotters approval, you both deny any wrongdoing. and you're of course watching d.w. news live from berlin up. next is our tech show shift. she just looks at why a speech recognition doesn't recognize everyone speech. on the spicer, thanks for watching. the literature invites us to see people in particular like to see as the fine might work if you do the books on you to the fight against the corona virus. pandemic is the rate of infection been developing. pleasures are being taken

just the latest research says information and context. coronaviruses special, monday to friday on w. well, speech be the new intersection between humans and technology. even twitter is setting its hopes on audio messaging. now, even twitter setting its hopes on audio messaging. now, how well does small voice recognition work, and what else does our voice reveal about us? voice, command and voice analysis. our topic today on shit and asking the fridge for recipes, talking to your g.p.s. or turning off the lights with your voice. turn off the lights,

digital assistance using voice command, infiltrating our day to day lives, but why always so keen on them canis is a fan and his berlin by start up, the oval has developed an open source platform that anyone can use to program voice apps, which work on all devices. that's a huge part of us and that's what i find fascinating about. voice is how natural this type of interaction is when she comes in, but even a 3 year old child can ask for a specific song, although i can't read it all to, i'm not going outside to tell, you know, and then there's the older generation who never quite felt understood by technology you yet still, i know all of a sudden they see more free interacting with tech. and this had never quite worked before. this example shows just how much it's already possible. the 2 people don't speak the same language, but they can still communicate thanks to smart speech software. the state was going to let both more settle and visit a reporter. i'm looking for

a place to have lunch before going to the airport. there's a great place around the corner. you can take the train from there to be your point . are you going to better that wait, that a lesbian now with is the might it is the i, yes, they laid up way to go. but it's not as easy as the advertisement might have you believe. i know this from my own family. my father speaks german with an accent, and the speech recognition technology of the remote control often struggles to understand his commands. this has to do with who developed the algorithms and crucially with who didn't pashto on the speech impediment dialects, different cultural backgrounds. there are many people who don't sound like the average silicon valley coder of developers who come yet i cling and there are author comprehension problems. it's all ties out. also just from differences between genders does oft, women are frequently understood less well than men. for example,

these differences are largely due to the data used to train voice algorithms. the typical software developer is a white man who speaks english. in the beginning, developers use their own voices to train the voice assistance, a circumstance which still has consequences to this day. no voice assistant, whether produced by amazon, google, apple, i.b.m. or microsoft will understand every word but the error rates fall notable patterns. a study by stanford university, for example, found a 35 percent error rate for african-americans, as opposed to 19 percent for white americans, even though english is the standard language for voice systems. many languages can't be used at all. just 8 work across platforms for amazon, google and apple. google smartphone assistant was the most ever seen. it speaks 44 languages, but generally the rule for all devices is minorities are not a certain time or take minutes. if you lend and in many countries,

all there's no speech software, yes off because the demand just isn't worth investing in at the most out on the hill. so speech recognition systems don't work the same for everyone. plus, let's not forget that there are people with speech impediments and people who cannot speak or hear. a german research project has developed the 1st tour that enables speech software to recognise language. research is also focusing on going beyond the strict sense of for its meanings. after all, it's not just what someone says, but also how they say it that plays a big role in comprehension. this is especially intriguing for companies, as they hope to find out more about their customers. i'm in a great time today, you know, it takes only a few words to hear what kind of mood a person is and whether they're happy for angry. this ai program can filter out 6000 parameters from someone's voice, including pitch and tambor. developers from the start of my dear in munich,

are hoping that their smart voice analysis will enable them to sneak a peek into people's hearts. the software has already been tested in call centers, where it has been used something coming calls to analyze the callers, as well as the staff members of the book is this useless distance for control of the office consumer using the software, i felt as if i were automatically looking at things from the outside modest when i was and i thought about the customer a lot more happy customers are loyal customers and companies want to keep it that way. but they're also critical voices. increasing creating voice based profiles is problematic when you get to lecture and digital ethics. and his outburst can reveal a lot about my mood or my house phone bill was forced to call. those are very sensitive private status, and when they should be of no concern to companies, then at least not so long as i have not given my explicit consent and it's ok. but that's what companies want to find out as much as possible about their customers

from their voices. since 2800 amazon has a patent on effort, ties meant for people with illnesses or conditions. in other words, alexa is supposed to tell from a person's voice whether they're ill, or even depressive, and then recommend products accordingly. but how is it that this technology can work to speak? we need our entire body, that longs the trachea and the larrys with its vocal cords. in total more than 100 muscles, a part of creating sound and several parts of the brain too. and that's the reason why our voice can reveal important information about illnesses like parkinson's compared to a healthy person's voice of god. someone with parkinson's voice with some choppy. 7 young shoulders as a computer scientist, just me my working on detecting parkinson's disease and it's very early stage. you

know, you can are concerned there's a neurodegenerative disorders at which specifically attacks the motor functions on the left. or such, you can be detected some very early on in the mind muscle items which we have here . if you generate sounds of voices and present in the future. it may also be possible to detect other conditions like autism, depression, or currently called 94 through voice analysis. incredible, what's become possible with artificial intelligence and digital voice analysis. but it's worth bearing in mind that this technology also comes with wrists back in 2017 . the canadian startup lie a bird called sister with manipulated voices of politicians like donald trump and barack obama. i am not a robot. my engine nation is always different. i'm not a robot. my information is always different. well, that doesn't really sound bad real. is that why a bird has since then been bought and now it's part of a software for transcribing and creating pod casts. the integrated ai makes it

possible to edit the voice recordings. after you have finished recording like this, for example, i should probably get in shape this year. i've changed this quickly. i should probably watch t.w. ship this year. that sounds pretty good already, but it only works on line and so far, only in english, but the technology is developing rapidly. take the example of this french company, which is already working on manipulating voices in real time. french software company candy for some has specialized in cloning voices. not to take me home or love, what you're doing. our technology can synthesize a boys with just 80 words and then let that person's voice say whatever it wants to play. the porter to demonstrate the way the software works, the voice experts need samples, spine kind doctors, football team achieve the expected results. in 500 sentences contain all the

typical sounds german defense, and the eclipse ends in the early afternoon hours in southeast china. an algorithm disassembles the recordings of the small fragments. the meaning of the words is completely irrelevant. here. what the program is after is the sound of the individual fragments so that it has numerous sound bites from both speakers. now the algorithm can start comparing the fragments, thereby also comparing the tambour of both voices. the more words are given to the algorithm, the better it can train on its own. once training is complete, the software should be able to convert every sentence spoken by the male voice into the female voice and vice versa. but how well does the software work? can voices be manipulated live, and i'm already on a 29 years old and i'm from germany. i'm speaking in max's voice and it sounds very real at this conference mission, not. i was pretty close to what i ordered was sounds like

a can of and the other way around. i'm max, but i'm speaking an area known as voice right now. that's pretty convincing. but also quite a weird feeling. pretty impressive, but it doesn't sound good enough to really be able to deceive you. scientists are working on using the human voice as a type of password. the idea behind that is that register customers would be immediately recognized by their voice when they call a growing number of companies, including european banks. what for a place, identity verification for security questions, fingerprints and pins with voice authentication? that's a problem here for instituting dobbs dot in southwest germany. project head on is testing such a voice i.d. system minded. my voice is my password to a company. for instance, a bank has his voice samples just as they would for

a customer. the company software created a biometric profile from his voice sample with more than 100, identifying characteristics such as pitch or rhythm. and she knew when logging in via voice, the program compares the voice speaking with the stored profile. if they match, the customer's granted access. but what if somebody uses a digitally manipulated voice key liar, birds or similar programs have not yet reached a level of protection to be able to accurately recreate a hokum melody and all the characteristics of a human being's voice. but not all experts are convinced voice verification systems are safe to err. i see voice as an interesting addition. it's just, it's not just concepts, it's interest. now i wouldn't use my voice for banking and so on and survive pretty actually, and have to not just, the technology is only just getting started on the home. so at their rest doubts.

but despite these, the tech giants seem voice, command and analysis as the next major intersection between humans and machines. and the technology has already made it into our homes. researchers expect that before long our whole apartment or house will be equipped with microphones and speakers so that smart speakers might even become redundant. i have to say that i have some concerns about this. to mean my apartment is generally a really private place, and i don't want to feel like someone or something is listening to me when i'm at home. but what do you think about this? what do you use voice command for? let us know on you tube or facebook. i buy and see you soon.

lose hamilton. times formula one champion. only bring down talent spirit. we know this is a scary time for us. the coronavirus is changing the world changing. so please take care of yourself. keep your distance if you can stay and how we deal with you for here for you. we are working tirelessly to keep you

informed over platforms. we're all embarrassed to get on together and we'll make it through. stay safe, everybody. stay safe, stay safe, is released. stay safe all gotten out of them and you have a little boy and hadn't been moved out of bed with plenty. we are living.