China’s two most international web giants, Alibaba and Baidu, have both announced embedded voice software and services similar to those of Amazon Alexa and Google Home, indicating the significant growth potential of the market for natural language applications backed up by machine learning.
Alibaba announced Tmall Genie, based on a Mediatek processor and similar to the Amazon Echo. Rival Baidu announced two- and four-microphone far-field reference designs for its DuerOS software using Mediatek and Conexant chips, as well as support for Nvidia’s Shield TV streaming device.
The growth in this market, particularly the voice recognition and processing elements, will be a boon to certain chip providers. The Echo, which has shipped an estimated 9m units in the US and UK, runs on a chipset from Texas Instruments, but according to EETimes, the retailer is planning to add another semiconductor supplier to its OEM program, which also includes Conexant, Microsemi and NXP.
A boost from smart speakers, and other emerging voice/AI form factors, will be particularly welcome to Conexant, which entered Chapter XI bankruptcy protection in 2013. It emerged successfully on the back of mixed-signal chips which support voice control, and its leadership in that area has landed it the Baidu deal and opened up the colossal Chinese market.
This will be one reason why touch and biometric chip specialist Synaptics snapped it up last month for $300m, along with Marvell’s multimedia business for another $95m. The plan seems to be to make a powerhouse of human interface chips, building on an existing strong relationship with Apple as well as targeting China.
Of course a supplier relationship with Apple is a double-edged sword – we have seen what happened to Imagination (and Qualcomm) once Apple turned unfriendly. But these new purchases will make Synaptics more useful to Apple, especially with talk of the US giant preparing an iPhone that recognizes your face before it gives you access. Synaptics might be a natural company to benefit from that.
But if you have to share a bed with Apple some of the time, it’s great to be able to pop next door to a big Chinese firm like Baidu in order to hedge your bets.
Conexant got into the voice processing market in 2012, and the company’s AudioSmart brand of voice input processors and embedded far-field processing software has become adopted by CE device manufacturers in numerous products ranging from AI digital assistant devices and smart speakers to voice-enabled televisions and personal robots. LG Electronics has integrated a Conexant voice input processor into its smart home products and Qualcomm has used it for its Hexagon digital signal processor family.
Last December Conexant cut a deal with Amazon to work with Amazon Voice Services, and then cut the time to market for Alexa, already the fastest selling voice input device in the world.
Now Conexant looks ready to begin powering Chinese domestic devices with two key advantages – superior ability to deal with noise and echoes when a voice needs to be understood in a noisy environment (like when music is playing); and to locate where that voice is coming from, so the device can listen more intently in that direction.
“The voice revolution is a global phenomenon. By working with Baidu we help more third party manufacturers bring to market innovative voice-enabled AI devices with an exceptional conversational AI experience,” said Saleel Awsare, president of Conexant. “The launch of Baidu’s DuerOS development kits and reference designs will drastically reduce development time and cost, allowing manufacturers to quickly bring their innovative ideas to market.”