Bravo
If ARM was an arm, BRN would be its biceps💪!
Was tryiing to drill down a bit more on the "self-learning" capabilities of Bosch's BHI260AP sensor as it states "When engaged, this system can be used to learn a new activity that it has never seen before" and private data is being "processed at the sensor instead of the main controller".
Also, Qeexo's AutoML can be deployed on Bosch's BHI260AP sensor.
Recently, BOSCH announced its latest activity sensor, the BHI260AP, but this sensor is more intelligent than most. What features does the BHI260AP integrate, what can its AI core do, and how does this sensor demonstrate the growing importance of AI?
The first AI was only ever found on powerful computer systems due to the large processing and memory requirements. Still, improvements in technology now see AI systems readily working on desktop PCs, smartphones, and even IoT devices. However, if the functions of AI can be shifted away from processors and onto dedicated hardware, then the main processor can free up its resources, thereby improving the device's capabilities in question.
This need for a “co-processor” is now seeing the development of neural network accelerators that utilise technology similar to that found on GPUs. Such hardware not only removes the AI functions away from the main processor, but are specifically designed to execute neural networks, and thus can do so more efficiently.
For example, the BHI260AP can be used for pedestrian dead reckoning whereby the low-powered sensor can determine where a user has walked in-between GPS cycling. Since GPS is a power-intensive operation, the use of power cycling reduces overall power consumption. The BHI260AP can also be programmed to perform relative and absolute orientation using the internal IMU, which can be useful for positioning devices.
The BHI260AP also integrates a range of pre-programmed activities which it can recognise, including jogging, jumping, and swimming. When used on a wristband, the BHI260AP can determine the number of steps, length of a stroke, or the wearer's swimming style.
The build-in AI mechanism can detect how an activity works if that activity involves a repetitive cyclic motion. This clearly demonstrates how sensors themselves are now becoming small data processors in their own right. Furthermore, this also demonstrates how potentially private data (position, acceleration etc.) is being processed at the sensor instead of the main controller. While the hose CPU can obtain raw sensor data, it is worth noting that future generations of sensors may hide this information, and provide a sanitised value instead.
"The self-learning AI sensor will change how users interact with their fitness devices from a mere one-way approach to an interactive way of training. This new sensor combines Bosch Sensortec’s long-term experience in smart motion sensors with its strong competence in innovative software development."
www.prweb.com
Also, Qeexo's AutoML can be deployed on Bosch's BHI260AP sensor.
BOSCH BHI260AP Self-Learning AI Sensor
06-01-2021 | By Robin Mitchell
Recently, BOSCH announced its latest activity sensor, the BHI260AP, but this sensor is more intelligent than most. What features does the BHI260AP integrate, what can its AI core do, and how does this sensor demonstrate the growing importance of AI?
How AI is Evolving
Artificial Intelligence, or AI, has come leap and bounds from the initial development of neural nets to software services that can learn as they operate. Unlike traditional software programs, AI has the advantage that they can recognise patterns more easily than a combination of IF statements and can be retrained to improve their performance over time.The first AI was only ever found on powerful computer systems due to the large processing and memory requirements. Still, improvements in technology now see AI systems readily working on desktop PCs, smartphones, and even IoT devices. However, if the functions of AI can be shifted away from processors and onto dedicated hardware, then the main processor can free up its resources, thereby improving the device's capabilities in question.
This need for a “co-processor” is now seeing the development of neural network accelerators that utilise technology similar to that found on GPUs. Such hardware not only removes the AI functions away from the main processor, but are specifically designed to execute neural networks, and thus can do so more efficiently.
BOSCH Releases the BHI260AP
Recently, BOSCH announced the release of their latest activity sensor, the BHI260AP, that integrates many features needed for activities (i.e. sports), including a 3-axis accelerometer, 3-axis gyroscope, 32-bit ARM EM4 programmable microcontroller, and a 4-channel micro DMA controller. The BHI260AP integrates a range of software features that can be useful for offloading tasks from the main controller.For example, the BHI260AP can be used for pedestrian dead reckoning whereby the low-powered sensor can determine where a user has walked in-between GPS cycling. Since GPS is a power-intensive operation, the use of power cycling reduces overall power consumption. The BHI260AP can also be programmed to perform relative and absolute orientation using the internal IMU, which can be useful for positioning devices.
The BHI260AP also integrates a range of pre-programmed activities which it can recognise, including jogging, jumping, and swimming. When used on a wristband, the BHI260AP can determine the number of steps, length of a stroke, or the wearer's swimming style.
BHI260AP AI Learning
However, by far the most impressive feature of the BHI260AP is its built-in AI self-learning mechanism. When engaged, this system can be used to learn a new activity that it has never seen before. A demonstration by BOSCH shows a user engaging the learning feature via a smartphone app and performs a few exercises (skiing in this case). Then the sensor can determine the number of strokes during skiing.
The build-in AI mechanism can detect how an activity works if that activity involves a repetitive cyclic motion. This clearly demonstrates how sensors themselves are now becoming small data processors in their own right. Furthermore, this also demonstrates how potentially private data (position, acceleration etc.) is being processed at the sensor instead of the main controller. While the hose CPU can obtain raw sensor data, it is worth noting that future generations of sensors may hide this information, and provide a sanitised value instead.
"The self-learning AI sensor will change how users interact with their fitness devices from a mere one-way approach to an interactive way of training. This new sensor combines Bosch Sensortec’s long-term experience in smart motion sensors with its strong competence in innovative software development."
- Dr. Stefan Finkbeiner, CEO at Bosch Sensortec
Qeexo, and Bosch Enable Developers to Quickly Build and Deploy Machine-Learning Algorithms to Bosch AI-Enabled Sensors
/PRNewswire-PRWeb/ -- Qeexo, developer of the Qeexo AutoML, and Bosch Sensortec GmbH, a technology leader in MEMS sensing solutions, today announced that...
