By Anne Morris
Qualcomm sets sights on carving out a role in voice activation
Creating the underlying technology that can support the next-generation of voice-activated products is complex and requires expertise in areas like compute, connectivity, sound quality and AI, noted Chris Havell, head of consumer/Bluetooth audio at Qualcomm Technologies’ voice and music business unit.
“These are all areas we excel at,” Havell said. Indeed, Qualcomm is already focussing on making voice activation and assistance experiences better for end-users, while also bringing these improved voice activation features to a wide array of consumer devices – from smart headsets to smart soundbars.
Havell said this has put huge focus and investment on underlining technologies and algorithms to capture voice cleanly, accurately identify when the “wake word” has been said by the actual product user, get the voice utterance into the cloud and interact with a multitude of different ecosystems in regard to how the response is handled.
“The wake word recognition aspect is a critical gate, and we have built up specific engineering resources, compute power and databases to be able to accommodate recognition of a broad range of wake words in different languages and accents,” he said.
Havell singled out consumer audio as one of the sectors where voice activation is in highest demand. “Take smart speakers for example. These products straddle the worlds of smart home control, consumer audio and smart assistance and their extremely fast uptake is helping drive the demand for voice UI more broadly,” he said.
Havell also pointed out that voice control has now become a must-have feature in most other consumer audio products, including headsets and soundbars, “and we already have the technologies needed to make these smarter audio devices a reality”.
Other sectors where he expects voice activation to play a big role include automotive, IoT and industrial IoT. But developing voice-activated products is challenging and requires complex integration and configuration, as well as power optimisation.
“Systems need high levels of processing in order to support always-on voice activation, as well as other rich features, but they also need to be power efficient enough to last for hours of standby time and playback without loss of functionality,” said Havell. He also cited actual voice capture and integration of multiple connectivity technologies such as bluetooth, Wi-Fi and Zigbee on the same device, as among the biggest challenges of developing systems based on voice activation.