Processing audio and voice on the IA8201 keeps the power-hungry host processor off as long as possible to extend battery life. According to Knowles, the IA8201 delivers better far field voice understanding and processes commands more efficiently in noisy, real-world environments. In addition, it enables new audio use cases beyond what the host processor provides, and through concurrent sensing of voice, audio and other sensor inputs, delivers a more natural user experience.
“Today’s consumers expect better listening accuracy in their electronics, which drives systems that have more mics and more processing. Device makers are also highly motivated to enhance user privacy and keep more personal information on-device. Together, this creates huge challenges for designers to implement these enhanced capabilities locally versus in the cloud,” says Mike Polacek, President of Intelligent Audio at Knowles. “Knowles AISonic Audio Edge Processors provide advanced multi mic processing at 10 to 100 times the efficiency of generic processors, which helps our customers design compelling products with performance and privacy features that end users demand.”
Sampling now, the IA8201 audio processor packs 1.44MB of memory, numerous interface options, and dual 175MHz cores with voltage scaling in a tiny 2.6x3mm eWLB package and a 6.0x6.0mm QFN package. Low power is achieved with a combination of cores offering the best combination of optimized instructions and industry compatibility.
The IA8201 includes a high compute 128-bit core (DMX) with Knowles proprietary instruction set and a Tensilica HiFi3 core (HMD), both with Knowles audio enhancements. The DMX is a 4-way floating-point SIMD processor targeted towards efficient performance computing (e.g. beam-forming, barge-in, AEC), while the HMD is targeted towards efficient, low-power, wake-on-voice applications with a two-way floating-point SIMD processor. Both cores contain dedicated accelerators for FFT, peak finding and DNNs.
The IA8201 solution provides generous memory (1.4MB) that is sized for specific use cases and architected for single-cycle access and low power. This large memory size can be used to store various keyword models, engines or other algorithms. Rich interface options include 2x I2C, 2x SPI, 2 x UART, and 24 GPIOs with interrupts to support various sensor and data connections, plus 4x PDM in, 2xPDM out, 3x I2S/TDM (4ch in/out) to support up to 4 microphones and various audio connections.
The IA8201 audio processor will be ready for mass production in Q3 2019. Knowles’ open DSP platform facilitates partnerships with third party algorithm providers who will be rolling out IA8201 compatible keyword models, various context detection plugins, and machine learning platforms including TensorFlowLite. The ROME framework which forms the foundation of the operating system is open to developers.
“With the launch of IA8201, Knowles has achieved high performance and low power in a cost effective solution, ideal for high volume smartphone, IoT and ear end-products. The IA8201’s audio optimized instruction set allowed us to implement a compact and power efficient solution in record time. The compact code footprint allowed us to pull in our speech technology features, previously executed on the application processor, to now run on Knowles optimized audio edge processor,” states Dr. Zhou Chief Technology Officer of AISpeech.
"The IA8201 is, simply put, the chip we've been waiting for. Customer requirements and expectations for sound quality are growing, requiring Alango advanced voice and audio processing algorithms combining reliable voice activity detection, multi-microphone beamforming, long-echo tail echo cancellation, sound personalization and efficient use of alternative bone conduction voice sensors. In many cases, the computational power, memory size and power consumption required to run all the required technologies simultaneously on the host processor or on an integrated Bluetooth chip is far beyond what can be allocated. The IA8201 allows offloading these advanced tasks while reducing the power consumption due to its complex numbers arithmetic, FFT and other accelerators. The IA8201 can be an enabling solution for years to come. Alango is committed to supporting Knowles’ customers using the IA8201 to make their products sound better," adds Dr. Alexander Goldin, Alango Technologies Founder & CEO.
“Consumers feel more comfortable with AI running on the edge, rather than everything being transmitted up to the cloud for analysis. The Knowles AISonic family of processors and especially the IA8201 provide a great solution for manufacturers who want to add the AI sense of hearing directly on a wide range of edge devices,” says Dr. Chris Mitchell, CEO and Founder of Audio Analytic.
"The partnership between Knowles and Elliptic Labs brings presence, proximity, and gesture detection to devices while delivering OEM benefits such as higher performance and lower power consumption. Knowles AISonic Audio Edge Processors are optimized for machine learning and works seamlessly with Elliptic Labs' Virtual Smart Sensor platform, which also uses machine learning and AI to deliver accuracy and precision,” reveals Laila Danielsen, CEO of Elliptic Labs.
"Our customers have increasing appetite for more MIPS and memory while maintaining power consumption that aligns with battery operated devices with wake-on-voice standby processing. After investigating several platforms, we found Knowles AISonic Audio Edge Processors to be the best solutions to meet these customer needs at a cost effective price-point. Our multi-channel acoustic echo-canceller was recently ported to the platform and the throughput was approximately ten times more efficient than on an Arm A53/NEON core, with no compromises in performance. The instruction set and floating-point support has enabled us to migrate from conceptual development to running it on the hardware with very little porting efforts. The two DSP cores together allow us to distribute processing in way that nicely balances computational efficiency vs power consumption," says Retune DSP's CEO, Thomas Andersen.
“Sensory and Knowles are both focused on high-performance artificial intelligence at the edge. By combining Sensory’s TrulyHandfree version 6.0 wake word technology with Knowles AISonic Audio Edge Processors’ multi-mic voice processing technology, customers can achieve high-performance, embedded voice solutions for smartphones, smart speakers, headphones and IoT devices,” states Todd Mozer, Sensory Chairman and CEO.
“Waves Maxx Technologies on Knowles’ DSP platforms embodies the perfect collaboration to deliver user-centric experience for the ‘Voice as UI’ revolution. With Knowles AISonic Audio Edge Processors for IoT and for ‘Voice in your Ear’ integrated with Waves’ technologies, OEMs who demand streamlined low-power consumption efficiency with optimal audio experience will have exactly what they need,” adds Tomer Elbaz, EVP of Waves Consumer Division.