Huami unveils chip for wearables, to be mass-produced in Q4

Huami Technology today hosted the first Huami AI Innovation Conference in Hefei and unveiled the second generation of smart wearable chip based on the RISC-V architecture: Yello Mountain 2.

According to Huami, Yello Mountain 2 has high computational efficiency, low power consumption, and the ability to implant a new generation of smart wearable devices. The NPU with convolutional neural network acceleration technology greatly improves the computational performance of local AI data.

It recognizes AF 7 times faster than in Yello Mountain 1, up to 26 times faster compared to pure software algorithms. 

Yello Mountain 2 has a built-in C2 co-processor, so even when the main chip is off, it's still possible to get a 26x increase in performance. The C2 co-processor is still able to record health data 24 hours a day.

This theoretically reduces power consumption by 50%, making it perhaps the world's best wearable chip.

Yello Mountain 2 has already been successfully streamed and is expected to go into mass production in the fourth quarter of this year.

In addition, Huami also announced several engine products at the conference.

This includes the big data-based motion engine ExerSense, which enables intelligent matching of motion models, and can currently identify the movement of the device. 19 motion modes covering 95% of the user's daily exercise routine.

The second generation heart rate engine, RealBeats 2, enables effective nighttime and daytime AF monitoring up to 1 generation 1.87 times and 6.64 times. It enables near-round-the-clock monitoring of AF and successful monitoring of refractory tachycardia and supraventricular premature beats.

There is also the oximetry engine, OxygenBeats, which pre-processes the oxygen signal with Huami's own biological data model. With a 50% improvement in accuracy over non-signal pre-processing equipment, and with multiple calibrations, the detection success rate has increased by 23%, compared to the previous year. Medical-grade oximeter with an average error of less than 1.67% will be available to everyone in Q3 this year.

The sleep engine SomnusCare accurately identifies the three sleep stages of light sleep, deep sleep, and REM sleep for over 25-minute nap data is detected with near 100% accuracy.

Combined with the OxygenBeats Oxygen Engine, it intelligently identifies sleep apnea syndrome and alerts you to Users take timely medical action.

Huami has also developed a new generation of high-precision optical sensors with additional support for oximetry engines and sleep engines BioTracker 2.