The 2017 Speech Industry Star Performers: Google
In the Smart Home Race, Google Gains on Amazon
The surge in demand for voice-activated speakers is sparking a new battle between the world’s major technology giants, with Google and Amazon vying to be the dominant computing platform in the connected home. And while Amazon’s Alexa holds the top spot right now, Strategy Analytics has predicted that Google Assistant will be able to leverage its advantages in search, machine learning, and ecosystem scale to become the leading platform.
“Google Assistant has the potential to dominate the intelligent home speaker market in the long term, assuming it can capitalize on its clear strengths in search and machine learning,” noted Bill Ablondi, director of Strategy Analytics’ Smart Home Strategies research program, in a statement.
Google continues to see increased adoption of its smart home assistant and a growing ecosystem of partners linking it to their own systems and applications. Mercedes, Hyundai, Sony, FirstAlert, Netflix, Samsung, Phillips, and Progressive Insurance are just some of the companies to integrate with Google Assistant in the past few months.
Google Assistant is a key component of the Google Home speaker, which the search engine giant released in Canada with a French language voice interface in late June. It marked the first foreign-language launch for the smart speaker. Google Home was released in the United Kingdom in May and has been available in the United States since last November.
Another unique feature that gives Google Assistant an advantage over some competitors is multiuser support, enabling it to provide personalized services based on users’ voices. Google rolled out the service—which allows Assistant to read off different calendars or remember music and news preferences for different people in the same house by recognizing their voices—earlier this year.
Google also partnered in the past year with both Dashbot and VoiceLabs to provide analytics services for Google Assistant developers. To further help developers, Google also released Actions on Google, a platform for building third-party voice-based personal assistant applications for Google Home.
Google Assistant is also benefiting from Google’s continued improvements in its speech systems. On the heels of its purchase of artificial intelligence start-up DeepMind in 2014 for $532 million, Google is launching technology that makes computer-generated speech sound more natural. The technology, called WaveNet, uses artificial intelligence and neural networking to learn how to form the individual sound waves that the human voice creates, thus enabling it to mimic human speech.
In a product sampling using English and Chinese speech, WaveNet was found to outperform Google’s previous speech synthesis systems by 50 percent.
To model waveforms, WaveNet takes 16,000 samples per second to generate data that it can then convert into speech. For each sample, it has to predict what the sound wave should look like, which the company says requires a lot of computation power.
“Building up samples one step at a time like this is computationally expensive, but we have found it essential for generating complex, realistic-sounding audio,” a Google executive wrote in a recent blog post.
WaveNet can generate breathing and mouth movements and identify the characteristics of different voices, including male and female. The same technology might even be used to synthesize other audio signals, such as music, and could even one day bring other elements, such as emotions or accents, into the mix.
But the real challenge for Google will be to reduce the cost and computational power requirements so that the product can become commercially viable.