An optimized implementation of iFLYTEK’s on-device neural network speech recognition software suite is available for CEVA’s audio/voice DSPs. The tightly-integrated solution is available for customers and is already in production for an ultra-low power voice processor targeting high volume consumer electronics.
Speech recognition is fast becoming the Human Machine Interface (HMI) of choice for consumer electronics, the smart home, mobile and wearable devices, surveillance, automotive and IoT in general, on the back of advances in sound processing and artificial intelligence. iFLYTEK is the leading speech recognition solution provider in China, and worked with CEVA to optimize iFLYTEK’s neural network based speech recognition, noise reduction and echo-cancellation algorithms for CEVA’s advanced audio/voice DSPs. The result is a highly-accurate, on-device voice processing solution, capable of enabling multiple mics voice activation without requiring cloud access.
“The combination of iFLYTEK’s software and our audio/voice DSP offers a powerful solution for embedding intelligent voice applications into mass market consumer electronic devices," said Ran Soffer, vice president of marketing and corporate development at CEVA.
Each CEVA sound processor is supported by hardware and software development tools, software libraries and the extensive CEVAnet partner ecosystem. http://www.ceva-dsp.com/app/audio-voice-and-speech/.
iFLYTEK claims to have over 70% of Chinese speech industry market share, providing core speech technology to over 2,000 companies. It launched the first "voice cloud" platform, which provides intelligent speech interaction capability for mobile internet industry and this has over 80,000 project partners with 700 million end-users in education, mobile phones, automotive, appliance and other industries, serving millions of households.