BHI260AP AI smart sensor integrates a 6-axis IMU, a 32-bit customizable programmable microcontroller, and various software functionalities. The AI smart sensor has embedded AI with on-sensor applications such as fitness tracking, navigation, machine learning analytics, and orientation estimation. The dimensions of the miniaturized AI smart sensor are 4.1mm x 3.6mm x 0.83 mm.
Hardware Features of BHI260AP AI Smart Sensor
- ARC EM4 CPU includes ARCv2 16/32 bit instruction set working up to a frequency of 3.6 MHz.
- The core also integrates Floating Point Unit (FPU) and Memory Protection Unit (MPU) with 4 channel micro DMA controller.
- CPU has two modes of AI functioning at 25Hz and 50Hz with 249µA and 386µA of current consumption respectively.
- Integrated sensor (6-DoF IMU) includes 16-bit 3 axis accelerometer and 16-bit 3 axis gyroscope.
- The sensor works at an operating voltage of 1.8 V with a standby current value of 8µA, hence the power consumption is only 0.014 mW in standby mode.
- Self Learning AI software has an activity learning time is less than 30 seconds. The recognition F1 score (the measurement of accuracy) is between 0.95 to 1.
- The host interface is configurable as SPI or I²C, 2 master interfaces (1 selectable SPI/I²C and 1 I²C), Up to 12 GPIOs
Software Features of BHI260AP AI Smart Sensor
- AI software for fitness tracking: This software allows on-device learning because of embedded AI. Additionally, it provides automated monitoring of various fitness movements, “including options for on-device individual-specific personalization of movements and support for increasing numbers of activities”
- Swim analytics: This software focuses on wrist wearables for the generation of essential information about users’ swimming activities, such as length count, style of swimming, and stroke counts. It also enables the detection of four different swimming styles: backstroke, butterfly, breaststroke, and freestyle.
- Pedestrian dead reckoning: This software enables power saving in wearable devices by activating the duty cycling of GNSS components. This also improves the accuracy of outdoor positioning with pedestrian dead reckoning.
- Relative and absolute orientation: This software evaluates the estimation of the relative and absolute orientation of the device. This includes the “outputs such as rotation vector, game rotation vector, linear acceleration, and gravity.”
AI Functionalities of the BHI260AP AI Smart Sensor
The AI smart sensor detects the motion of the user and recognizes the workout without the user mentioning it. The sensor can also perform unsupervised learning algorithms to detect the new workouts performed by the user, by clustering similar movements by pattern recognition. It also has a feature of letting the user know when the physical form of the workout is off.
The data can be directly communicated to a smartphone without the need to sending it through the cloud, because of the embedded AI. The power consumption is reduced as no additional network is required. Also, the data is secured and allows for continuous tracking and analysis.
For the development of AI-based wearable solutions, the company “offers an application board along with a BHI260AP shuttle board, which can be wirelessly connected over Bluetooth Low Energy (BLE) to smartphones.” The AI smart sensor has a software development kit for programming with additional custom coding which enables to program it for dedicated applications”.
Bosch Sensortec has an online community that can help you to solve your specific queries related to development. The BHI260AP AI smart sensor is available via Bosch Sensortec’s distributors. The different software packages are available for download at the Bosch Sensortec website. There is no information given about the application board on the official product page at this time.
The images and the device specifications have been taken from the official product page.
Saumitra Jagdale is a Backend Developer, Freelance Technical Author, Global AI Ambassador (SwissCognitive), Open-source Contributor in Python projects, Leader of Tensorflow Community India and Passionate AI/ML Enthusiast
So we’ve really reached the point where vendors try hard to write “AI” everywhere even where it is totally unrelated, just like they would have written “cloud” 10 years ago. They just paid an intern to develop an application supposed to connect to the device over bluetooth and surf on the AI wave.
Too bad they hide the real differentiators of their chip behind these marketing lies, as it seems the size and consumption make it a good product for a 6-axis IMU and integrated CPU+FPU.