Conexant Partners with DIOTEK for Speech Recognition in Embedded Applications

Conexant Systems announced a technology partnership with DIOTEK to integrate speech recognition with its  digital audio processor, the CX20805. The joint solution is targeted to smart home appliances, consumer electronics, and educational toys.

However, enabling voice commands for many products does not come without challenges. For home appliances, ambient noise and distance must be taken into consideration. Conexant has addressed these issues with its pre-processing algorithms that include noise reduction and far-field voice input processing. Conexant provides the hardware and advanced pre-processing algorithms to filter out ambient noise and provide a good hit-rate for DIOTEK's speech recognition engine to produce reliable voice control from far away.

Target areas of automatic speech recognition (ASR) segmentation include:

  • Voice Wake Speech Recognition - power-on scenario with no audio playback and magic word scenario with audio playback
  • Keyword Recognition - full ASR on-chip parameter feature extraction, lookup, recognition, and fixed words library
  • Natural Language - cloud-based recognition and response

"Our DSP with its integrated memory and patented far-field voice input processing technology coupled with DIOTEK's speech recognition engine enables us to offer our customers a low-power, high-performance solution to meet demands in a variety of applications," said Sailesh Chittipeddi,  president and CEO of Conexant, in a statement. "This powerful combination of technologies will not only produce crystal clear speech recognition but also reduce time to market for our customers."

"With the ever-increasing number of applications that will enable consumers to wake up their devices and control them with their voices, there is a great need to produce voice and speech recognition technology to meet the demand," said Jeong In Doh, CEO of DIOTEK. in the statement. "Our partnership with Conexant, an established leader in far-field voice input processing, will help to make true hands-free operation for a variety of products a reality."

The CX20805 from Conexant is a dual-core 32-bit, high-performance, low-power digital audio processor targeted to applications with the most demanding audio requirements. Based on the Conexant Audio Processing Engine (CAPE) architecture with scalable performance of up to 800 MIPS, the CX20805 integrates high-speed USB 2.0 and other key audio interfaces with advanced power management functionality.

SpeechTek Covers
for qualified subscribers
Subscribe Now Current Issue Past Issues