CEVA, a licensor of digital signal processing and IP platforms for cellular, multimedia, and connectivity, and Cypher, a provider of pattern recognition and audio enhancement solutions, have partnered to offer a noise reduction solution for smartphones, tablets, and wearables, and greatly enhances speech recognition performance in noisy environments. The solution integrates Cypher's voice-isolation software with the CEVA-TeakLite-4 DSP, suitable for any low-power application, including always-on' voice-activated devices.
Cypher's algorithms are based on a neural network analysis of human speech, which can tell human voices from other sound sources and isolate them. The voice isolation software can significantly improve the clarity and intelligibility of the speech, even in noisy environments.
"The breadth of applications for voice technology in recent years has expanded beyond handsets to include a whole new range of devices deploying voice activation and control," said John Walker, CEO of Cypher, in a statement. "Our patented voice isolation technology significantly boosts clarity as it eliminates over 99 percent of background noise and also improves speech recognition as it cuts the word error rate by 40 percent in noisy environments. The CEVA-TeakLite-4 DSP is fully equipped to handle any use case, delivering outstanding performance within a highly compact memory footprint."
"The CEVA-TeakLite family is the industry standard DSP for integrated audio and voice, having powered more than 4 billion devices to date," said Moshe Sheier, director of product marketing, audio and voice at CEVA, in a statement. "We are pleased to add Cypher's voice isolation solution to our rich ecosystem. Cypher brings a unique solution for enhancing voice calls and enables a whole new range of applications where voice activation technology can be utilized for the Internet of Things and beyond."