Qualcomm voice SoCs should make smart speakers out of all of us

A lot of consumer publications have been talking about Voice Assistants for a while now, but the development of them to date have been entirely proprietary, and they are given away for a price below their cost, and pushed to capture the market, before anyone clearly knows what to do with it. They rely on the existing natural language processing systems of a handful of companies, but also the core listening and music projection capabilities of speakers and microphones which could be made by anyone.

People instinctively know it is a market, that it will become a platform and that natural language processing, will become like the PC or the Smartphone, there will be one or two winners, who control the experience globally. But having a good language API is one thing, the device design that opens the door to that design is another thing entirely. So far, the thinking says that the voice API owner must make the device – we beg to disagree.

Right now Apple, Microsoft, Amazon and Google have the war to themselves and although you can buy a voice assistant for $50 (EchoDot), they certainly cost more than that to make, and the spend on building out an eco-system will make these loss leaders for some time. But Apple has virtually lost all hope of being an innovator again, Microsoft too, Google has misfired in many markets (Google+) and Amazon is quietly competent, and currently leading the charge.

Real innovation in this market will come from involving lots of smaller start-ups, which can chase down the revenue opportunities which are down every alleyway. That leads to 1,000s of voice applications, which in turn creates the eco-system enrichment needed for not just smart speakers, but for voice assistants, to take off. At the cloud end of this eco-system there is talk of AI, but for the most part the AI engines are also provided by Apple, Microsoft, Amazon and Google. While at the device end, this week Qualcomm has facilitated more companies getting into the device market by offering up its own chip platform to compete with the voice assistants from the majors out there – initially just for smart speakers.

This app is typified by Apple’s recent HomePod, which we slated as too expensive last week, which was unveiled at Apple’s WWDC. The issue is why does every speaker require a chip powerful enough to run a smartphone?

Qualcomm showed off its entry at its voice and music developer conference in Shenzhen in China this week, calling it the Qualcomm Smart Audio Platform, saying it brings together processing capability, connectivity options, voice user interfaces and premium audio technologies in two SoC variants plus software, so that OEMs can pile into this market. This is the type of move that will provide high volumes, a multiplicity of products and competition to drive down pricing.

The voice elements of the Qualcomm SoCs support multi-microphone, far-field voice, making highly responsive voice activation, as well as using beamforming technologies. The software also incorporates echo-cancellation, noise suppression and ‘barge-in’ capability. You still want your voice assistant to hear you even if your music is playing loudly and other people are singing along to it.

Qualcomm calls its multi-room audio technology AllPlay and said it supports whole-home audio music streaming. It has perhaps thought the underlying technology through further than Apple, and says speakers can be configured to play synchronized music across multiple rooms, different music in multiple speaker zones, and multi-channel wireless surround sound to multiple speakers in the same room. Apple just wants you to put separate $350 HomePods in every room.

Consumers can play music from streaming music services including hi-res 32-bit/192kHz content, local music libraries on mobile phones, or DLNA media servers. Moreover, the new platform supports Qualcomm aptX HD audio technology, which is designed to deliver 24-bit high-definition wireless audio over Bluetooth. At Apple Bluetooth might be thought to be a dirty word, but Apple is expected to support it in HomePod when it is finally delivered, but it was not mentioned at the launch.

Qualcomm is always fighting the same fight. Apple builds a chip for its own products, and that helps clarify many chip design decisions early. In effect it can offer a subset of the potential capabilities. Qualcomm has to deliver all of the potential for many other capabilities in many technical environments, and supports as many standards as it can. Which is why Apple always gets certain types of chips to market before Qualcomm.

Qualcomm also introduced in Shenzhen the CSRA68100, the company’s next-generation Bluetooth audio platform for premium wireless speakers and headphones offering four times the DSP processing power of its predecessor, along with a powerful 32-bit dedicated developer application processor.

So the smart speaker market, which will define what devices you talk to in your home, and lay down the landgrab for voice assistants – which are destined to live both in-home and on-phone – is now enabled with an SoC which can be used to push out 100s of devices, many of which will allow a choice of natural language API – say Alexa and get Google, Say Siri and get Apple, say Cortana and get Microsoft – which may prevent the voice eco-system becoming as proprietary as the common browser. Hands up who doesn’t want voice to become a battle ground purely for the big four? Not so loud my voice assistant can’t hear me.