Real-time voice-enhancement tech works with DSP design software

Synopsys, Inc. and Alango Technologies today announced that Alango Technologies’ Voice Enhancement Package (VEP) software has been optimized for Synopsys’ DesignWare® ARC® Data Fusion IP Subsystem. The VEP is a suite of real-time software DSP technologies designed for improving speech recognition performance in voice-controlled multimedia devices. The ARC Data Fusion IP Subsystem is a highly integrated hardware and software IP product that processes data from sensor and audio sources with minimal energy consumption. The combination of Alango’s optimized audio signal pre-processing software with the ARC Data Fusion IP Subsystem enables designers to rapidly integrate a complete, pre-verified hardware and software solution for voice and speech processing into a range of multimedia devices requiring minimal energy consumption.

“The increasing prevalence of voice activation and recognition in human-machine interfaces is requiring sophisticated speech-to-noise ratio enhancement technologies to enable highly accurate speech comprehension in noisy environments,” said Dr. Alexander Goldin, CEO of Alango Technologies. “Optimizing our Voice Enhancement Package for the ARC Data Fusion IP Subsystem gives SoC designers an ideal low-power solution enabling superior voice activation, recognition and control capabilities in their products.”

Alango’s VEP is a suite of real-time software DSP technologies that includes a multi-microphone beamforming array and stereo echo canceller to enhance far-field speech pick-up and response to “barge-in” commands. The VEP is a front-end solution that performs audio signal pre-processing prior to the automatic speech recognition (ASR) and key word recognition (KWR) engines. The VEP receives the microphones’ signals, applies acoustic echo cancellation, and then forms acoustic beams relative to the usage environment or “field.” These acoustic beams, which have much-improved signal-to-noise ratio (SNR) compared to the raw microphone signals, are then sent to the ASR engine for recognition. VEP processing significantly improves the ASR’s ability to recognize verbal commands.

The ARC Data Fusion IP Subsystem is a pre-verified hardware and software solution optimized for highly efficient DSP performance and ultra-low energy consumption. The subsystem processes data from digital and analog sources, offloading the host processor to enable more efficient data processing. The configurable IP subsystem, with its choice of EM DSP processors, delivers a 2x performance boost for typical signal processing functions compared to other available processors and significantly reduces frequency and memory requirements to run audio codecs and speech/voice communication software. The ARC Data Fusion IP Subsystem also includes an audio processing library of common functions, including gain control, mixer and sample rate conversion. Tightly coupled PDM and I2S peripherals simplify the integration of voice and audio I/O such as MEMS microphones used for far-field voice user interfaces and hands-free voice commands. In addition, the hardware PDM interface implementation is significantly more energy efficient than the equivalent software implementation, with minimal gate count impact. The integrated solution is optimized for “always on” data fusion combining sensor, voice, gesture and audio processing functions.

“Advanced IoT devices are incorporating multiple microphones to enable more precise voice recognition by identifying the origin of a sound and cancelling interferences,” said John Koeter, vice president of marketing at Synopsys. “By combining our ARC Data Fusion IP Subsystem with Alango’s Voice Enhancement Package, designers can quickly integrate high-quality voice and speech capabilities into their SoCs while staying within their power and area budget.”