{"id":11895,"date":"2020-11-02T15:24:36","date_gmt":"2020-11-02T14:24:36","guid":{"rendered":"https:\/\/www.codemotion.com\/magazine\/?p=11895"},"modified":"2022-01-05T20:06:14","modified_gmt":"2022-01-05T19:06:14","slug":"voice-control","status":"publish","type":"post","link":"https:\/\/www.codemotion.com\/magazine\/voice-digital-assistants\/voice-control\/","title":{"rendered":"Voice Control: Building Your Voice Assistant"},"content":{"rendered":"\n\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t\t\t\tTable Of Contents\t\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t
  1. Speech recognition in the real world<\/a>
  2. The search for voice control<\/a>
  3. A practical voice control implementation<\/a>
  4. Going beyond simple voice recognition<\/a><\/ol>\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\n\n\n

    Voice control<\/span> was the stuff of science fiction throughout the 20th Century. But in the last two decades, voice control has entered the mainstream. Voice assistants like Siri<\/span> and Alexa<\/a> are embedded in home devices, headphones, and even cars. <\/p>\n\n\n\n

    But, how did we get to this point? What is the connection with machine learning<\/span><\/strong> at the network edge? And how can you create your own voice-activated edge device<\/span>? This article, part of a series on machine learning<\/span> at the edge, answers all these questions and more.<\/p>\n\n\n\n

    Speech recognition in the real world<\/h2>\n\n\n\n

    Voice control<\/span><\/strong> always fascinated futurologists and Sci-Fi authors alike. Back when it was first proposed, it must have seemed like a distant dream. But over the past decade, voice control has become routine and mainstream. This is thanks to a combination of factors<\/span>. <\/p>\n\n\n\n

    Advances in speech recognition<\/span><\/em> and natural language processing<\/span><\/em>, the availability of powerful computers<\/span> for machine learning<\/span>, and the growth in high-power edge-devices. Nowadays, we can see examples of voice control all around us. <\/p>\n\n\n\n

    Virtual voice assistants<\/h3>\n\n\n\n
    \"virtual<\/figure><\/div>\n\n\n\n

    Virtual assistants<\/strong>, such as Alexa<\/span>, Siri<\/span> and Google<\/span> Assistant<\/span>, have driven a huge take-up in voice control systems<\/span>. In essence, a virtual assistant listens to your instructions and acts on these. For instance<\/span>, you can ask it to play music, to tell you the weather, or to navigate you to your destination. <\/p>\n\n\n\n

    In general, these virtual assistants all work similarly. They require a suitable edge device<\/span> with network connectivity and a powerful backend<\/span>. Typically, the edge device<\/span> may be a smartphone, a smart speaker, or, increasingly, some other device<\/span> like a TV or pair of headphones. <\/p>\n\n\n\n

    The edge device<\/span> just tries to detect a \u201cwake word\u201d. It sends the voice message<\/span> to the backend<\/span> for processing<\/span> and then handles the result that gets returned. This process clearly depends on a good Internet connection<\/span>. However, recent improvements in edge technology mean more and more functionality<\/span> can be kept on the device<\/span>. <\/p>\n\n\n\n

    Cars<\/h3>\n\n\n\n

    Driver distraction is one of the biggest causes of deaths and injuries on our roads. As a result, car manufacturers have invested billions into driver aids designed to help reduce distractions. One of the most powerful is adding voice control to cars. This allows the driver to interact with the infotainment system in a totally hands-free<\/span> manner. <\/p>\n\n\n\n

    Unlike the voice assistants above, such systems<\/span> cannot rely on having network connectivity. As a result, all the voice recognition<\/span> and processing must be done within the edge device<\/span>.<\/p>\n\n\n\n

    The search for voice control<\/h2>\n\n\n\n

    As mentioned above, voice control grew out of advances in speech recognition<\/span> and NLP<\/span> along with increased computing power<\/span>. Creating functional voice control required computer scientists<\/span> to solve a number of problems. <\/p>\n\n\n\n

    Firstly, how do you record a person speaking and convert this into text? Next, how do you parse that text to extract the meaning? Finally, how do you work out the correct response? <\/p>\n\n\n\n

    These problems have interested computer scientists<\/span> since long before the invention of the modern computer<\/span>. Indeed, the idea of teaching computers<\/span> to understand humans dates back to the earliest days of computing<\/span>. Each of these problems required a different solution<\/p>\n\n\n\n

    Speech recognition<\/h3>\n\n\n\n

    In the 1990s, more and more people got access to computers<\/span> at work and at home. Few of these people were able to type though. So, a lot of effort was invested in creating systems<\/span> that would allow a human to dictate to a computer<\/span>. <\/p>\n\n\n\n

    This process of converting your voice into words<\/span> on the screen is known as speech to text or speech recognition<\/span>. <\/p>\n\n\n\n

    The earliest speech recognition<\/span> systems were created in the 1950s. They were able to distinguish single spoken digits. However, it took until the late 1960s until speech recognition<\/span> became a serious research area. By the 1970s, systems<\/span> were being developed that could recognise longer words<\/span> and even phrases.<\/p>\n\n\n\n

    The real breakthrough came with the application<\/span> of hidden Markov models to the problem. By 1987, this resulted in the Katz back-off model<\/a>, which allowed practical speech recognition<\/span> on computers<\/span> or specialised processors.<\/p>\n\n\n\n

    Throughout the next decade, systems<\/span> got better and better at speech recognition<\/span> across multiple languages. By the mid-1990s, the technology<\/span> had advanced sufficiently for companies to start selling commercial speech recognition<\/span> systems. <\/p>\n\n\n\n

    Early speech recognition<\/span> often relied on training the system<\/span> to recognise a single voice. This was achieved by asking the user<\/span> to read out a specific passage of text. This text included all the possible phonemes, parts of speech<\/span>, etc. to allow the system<\/span> to learn<\/span> that user<\/span>\u2019s voice. <\/p>\n\n\n\n

    More recently, we have seen speech recognition<\/span> applying machine learning<\/span> approaches to learn<\/span> to understand different accents. This avoids the classic problem where the computer<\/span> is unable to understand someone with a strong accent. Modern systems<\/span> are now able to understand multiple regional and national accents. <\/p>\n\n\n\n

    Natural Language Processing<\/h3>\n\n\n\n

    Of course, just being able to write down what a person says is not enough. Voice control also requires the computer<\/span> to understand what is said. This is a much harder problem known as natural language processing<\/span> or NLP<\/span> for short. Here, the computer<\/span> must learn what you actually meant. If you have learned a foreign language, you will know how hard this can be. <\/p>\n\n\n\n

    The problem is, human language depends on all sorts of factors<\/span>. Context, emotion, knowledge, and idiom all change the meaning of a sentence. Often, there are many ways to say the same thing, even for something as simple as making a phone call. \u201cI\u2019m phoning my parents\u201d, \u201cI\u2019m calling home\u201d, \u201cI must give my dad a call\u201d, etc. NLP<\/span> is the process of teaching a computer<\/span> about the structure and meaning of human language.<\/p>\n\n\n\n

    For decades, NLP<\/span> was a theoretical field. Computers<\/span> simply weren\u2019t powerful enough to solve the problem. Nowadays, computers<\/span> are getting better and better at it. This is largely down to improvements in machine learning<\/span>, especially deep learning<\/a>. The latest approaches combine several different machine learning<\/span> technologies<\/span>. <\/p>\n\n\n\n