What Makes Us Different
- Robust real-time acoustic interference and echo cancellation enables embedded voice in noisy environments
- Ultra low distortion voice capture enables lower ASR word error rate
- Scalable implementation with Xilinx programmable SoC or custom SoC
- Algorithm and AI acceleration compresses more intelligence on single chip solution
- Flexible algorithm enables non-uniform custom mic array to fit your product
- Complete embedded voice platform for easy evaluation and prototyping
Sound Capture Technology
Noise & Echo Cancellation
Other technologies tout their ability to cancel noise and echo from their own speaker. This is easier to do as you have access to the interfering reference signal. At Aaware, we go beyond traditional AEC and cancel interfering noise were we don\’t have the reference signal. This is much more difficult but gives our customers the ability to integrate without needing the interfering reference signal to adequately cancel enough noise to enable superior far field voice interfaces.
Source Detection & Localization
Detecting and localizing sources in the sound field is a critical step in finding what you’re looking for. The Aaware technology not only localizes and detects omni directionally, but separates the source speech with ultra-low distortion which enables a lower Word Error Rate (WER) in noisy environments from third party Automatic Speech Recognition (ASR), such as Picovoice and Sensory. We can also forward source localization data to down stream applications, such as video.
We set out to solve the problem of voice user interfaces that don’t work in noise and not being constrained to just a few microphones and a software only implementation. This led us to the amazing benefits of algorithmic acceleration leveraging the power of FPGA hardware. In doing so, we provide our customers with the best possible price/performance using less power and all integrated into one fully programmable SoC.
Flexible & Scalable Algorithm
Our algorithm is architected such that it can be re-configured to support up to 13 microphones, with the current development platform, and to support non uniform array sizes. Product teams aren’t stuck with a rigid microphone configuration and don’t have to upgrade to larger SoCs just to add a few microphones, making this a flexible fit for many products.
Scalable Processor Choices
Aaware gives you plenty of processor implementation choices with the Zynq®-7000 SoC and Zynq® UltraScale+™ MPSoC processor families from Xilinx. This includes automotive, industrial and commercial grade parts with embedded processors that scale from a single 32-bit ARM Cortex™-A9 core to 64-bit ARM® Cortex-A53 quad cores.
Complete Embedded VUI Platform
Aaware integrates the entire voice stack, with partner wake words, speech recognition, natural language understanding, microphone array and presents a complete reference design in the form of demos all running within an open Ubuntu linux environment enabling a software-ready platform.