Intel To Add DSP, Speach Recognition To Upcoming Chip
After AMD's move to integrate DSP hardware into some of its latest Radeon graphics cards, Intel seems to be doing something similar with its 14-nm Broadwell processors, which will reportedly have an integrated DSP charged with powering Smart Sound Technology.
But Intel's approach is different than AMD's TrueAudio tech, which is designed to improve 3D audio and offer surround-sound virtualization for games.
Slides posted onlibe on VR-zone.com show that Intel's 'Smart Sound DSP ' will handle real-time mixing, decoding, and post processing. It will also support speech recognition without increasing the CPU draw, as today's speach recognition approaches such as Apple's Siri and do.
The report also mentions that the DSP will be able to play audio even when the system is in Windows 8's low-power Connected Standby mode. In addition, the Broadwell DSP's Connected Standby functionality extends to recording, allowing it to listen for voice commands while the system is sleeping. Intel's "Genie" personal assistant software will take advantage of this capability and will respond to the user's commands, the report claims. The assistant software is also said to be smart enough to recognize a specific user's voice.
Slides posted onlibe on VR-zone.com show that Intel's 'Smart Sound DSP ' will handle real-time mixing, decoding, and post processing. It will also support speech recognition without increasing the CPU draw, as today's speach recognition approaches such as Apple's Siri and do.
The report also mentions that the DSP will be able to play audio even when the system is in Windows 8's low-power Connected Standby mode. In addition, the Broadwell DSP's Connected Standby functionality extends to recording, allowing it to listen for voice commands while the system is sleeping. Intel's "Genie" personal assistant software will take advantage of this capability and will respond to the user's commands, the report claims. The assistant software is also said to be smart enough to recognize a specific user's voice.