MEMS Sensors Ecosystem for Machine Learning
The integration of artificial intelligence (AI) algorithms into MEMS sensors is transforming the way that we interact with the world around us. By embedding AI technology at the edge, today's sensors can collect, process, and send meaningful data in real time.
We enable the transition to in-sensor processing with a new generation of smart, open, and accurate sensors to help developers exploit their potential while improving overall system efficiency.
What makes our sensors unique?
- Smart sensors enable AI at the edge, reducing system data transfer volumes and offloading network processing for lower power consumption and a more sustainable solution
- An open ecosystem accelerates innovation and product development thanks to data sharing
- Accurate sensors bring meaningful information to the end users through the development of highly complex algorithms
To ensure developers find the most effective solution in terms of computing capacity and flexibility in programming, ST offers a choice of several technologies for in-sensor processing: sensors with an embedded machine learning core (MLC) and sensors with an intelligent sensor processing unit (ISPU).
Sensors with an embedded machine learning core
An MLC is an engine that can be trained to trigger an action when a specific event is detected using decision tree learning. With an MLC embedded in the sensor, it’s possible to recognize precise movements and communicate the event to a processor with the best possible system energy efficiency.
Added value:
- Extremely low-power solution
- Increased accuracy with better context detectability
- Offloads the main processor, improving system efficiency
MEMS sensors with an embedded MLC
Our third-generation of MEMS sensors with advanced machine learning core technology enables intuitive and context-aware functions for the latest battery-operated applications.
| Part number | Application | Family | MLC | Full scale | Temperature range | Power consumption | Application note for MLC features |
| LIS2DUX12 | Consumer | Accelerometer | 128 nodes | ±16 g | -40°C to +85°C | 2.7 µA | AN5903 |
| LIS2DUXS12 | Consumer | Accelerometer | 128 nodes | ±16 g | -40°C to +85°C | 2.7 µA | AN5901 |
| LSM6DSV16X | Consumer | iNEMO | 128 nodes | ±4000 dps, ±16 g | -40°C to +85°C | 0.65 mA combo | AN5804 |
| LSM6DSV16BX | Consumer | iNEMO | 128 nodes | ±4000 dps, ±16 g | -40°C to +85°C | 0.95 mA combo | AN5892 |
| LSM6DSV32X | Consumer | iNEMO | 128 nodes | ±4000 dps, ±32 g | -40°C to +85°C | 0.65 mA combo | AN6071 |
| ASM330LHB | Automotive | iNEMO | 512 nodes | ±4000 dps, ±16 g | -40°C to +105°C | 0.8 mA combo | AN5915 |
| ASM330LHBG1 | Automotive | iNEMO | 512 nodes | ±4000 dps, ±16 g | -40°C to +125°C | 0.8 mA combo | AN6068 |
| ASM330LHHXG1 | Automotive | iNEMO | 512 nodes | ±4000 dps, ±16 g | -40°C to +125°C | 0.8 mA combo | AN5987 |
| ISM330BX | Industrial | iNEMO | 128 nodes | ±4000 dps, ±8 g | -40°C to +85°C | 0.6 mA combo | AN6124 |
Other MEMS sensors with an embedded MLC
| Part number | Application | Family | MLC | Full scale | Temperature range | Power consumption | Application note for MLC features |
| LSM6DSOX | Consumer | iNEMO | 256 nodes | ±2000 dps, ±16 g | -40°C to +85°C | 0.55 mA combo | AN5259 |
| LSM6DSO32X | Consumer | iNEMO | 256 nodes | ± 2000 dps; ± 32 g | -40°C to +85°C | 0.55 mA combo | AN5656 |
| LSM6DSRX | Consumer | iNEMO | 512 nodes | ±4000 dps, ±16 g | -40°C to +85°C | 1.2 mA combo | AN5393 |
| ISM330DHCX | Industrial | iNEMO | 512 nodes | ±4000 dps, ±16 g | -40°C to +105°C | 1.2 mA combo | AN5392 |
| IIS2ICLX | Industrial | Accelerometer | 512 nodes | ±3 g | -40°C to +105°C | 0.42 mA | AN5536 |
| ASM330LHHX | Automotive | iNEMO | 512 nodes | ±4000 dps, ±16 g | -40°C to +105°C | 0.8 mA combo | AN5781 |
How to get started with sensors with an embedded MLC?
The best way to get started with machine learning core in sensors is to select the appropriate solution with supporting ST tools and software for your application.
Sensors with an intelligent sensor processing unit (ISPU)
The ISPU is a true integrated digital signal processor (DSP) that is optimized with respect to a general-purpose MCU and can be used to run complex AI algorithms. Its advantage is that - being integrated - it optimizes the required computing power to the maximum.
Added value:
- Ultra-low power consumption at system level, thanks to optimized data transfer
- High-processing capability with AI-enabled programmable core (ML and NN)
- Easily programmable with C language or with commercial and open-source AI models
MEMS sensors embedding an intelligent sensor processing unit (ISPU)
| Part number | Application | Family | Memory | Full scale | Temperature range | Power consumption | Application notes for ISPU features |
| ISM330IS | Industrial | Inertial Measurement Unit | 10 MHz clock, RAM 40 KB | ± 2000 dps, ± 16 g | -40°C to +85°C | 0.59 mA (combo mode) | AN5850 |
| ISM330ISN | Industrial (anomaly detection) | Inertial Measurement Unit | 10 MHz clock, RAM 40 KB | ± 2000 dps, ± 16 g | -40°C to +85°C | 0.59 mA (combo mode) | |
| LSM6DSO16IS | Consumer | Inertial Measurement Unit | 10 MHz clock, RAM 40 KB | ± 2000 dps, ± 16 g | -40°C to +85°C | 0.59 mA (combo mode) | AN5799 |
How to get started with sensors with an embedded ISPU?
How to program the ISPU?
There is an AI solution for every need using ISPU.
With MEMS Studio you can bring your own neural network model (built with Keras, TensorFlowLite, ONNX) into optimized C code for ISPU
ISPU-Toolchain (C complier)
We provide ISPU programming support with an ecosystem of libraries and third-party tools/IDEs to help you implement even the most complex AI models.
Embedded developers without any data science skills can use NanoEdge AI Studio to program the ISPU (ISM330ISN). You can readily obtain accurate intelligence solutions with a limited amount of time and effort.
Neuton.AI (from ST partner)
No-code TinyML platform enabling everyone regardless of experience or expertise to build and deploy Machine Learning models directly to an ISPU and/or to any MCU natively.
Recommended resources
Ready-to-go application examples in GitHub for AI at the edge
In our GitHub repository you will find application examples both for MLC and ISPU, such as human activity recognition, head gestures, vibration monitoring for predictive maintenance and more. To get started quickly with each example, the README file provides detailed information.
Webinars
| Event | Target |
| An intelligent sensor for sustainable always-aware applications | Machine learning core |
| In-sensor monitoring with intelligent MEMS sensors | Intelligent sensor processing unit |
| Anyone can build smarter applications with this intelligent IMU | Machine learning core |
| Predictive maintenance with AI at the edge in MEMS sensors | Machine learning core |
| AI for asset tracking using only machine learning core in sensors | Machine learning core |
| Implementing AI in sensors to develop power-efficient personal electronics applications | Machine learning core |
| Program decision tree in sensors with a Machine Learning Core | Machine learning core |
| Measure and protect! Introducing the LSM6DSV32X IMU | Machine learning core |
| Video | Target |
| ISPU: How to build your first example with Eclipse-based IDE? | Intelligent sensor processing unit |
| ISPU: How to program your own example starting from an ISPU template? | Intelligent sensor processing unit |
| ISPU: How to build a visual output of your project? | Intelligent sensor processing unit |
| MEMS sensor with AI core (ISPU – intelligent sensor processing unit) | Intelligent sensor processing unit |
| How sensors with a machine learning core bring power-efficient in AI to the edge | Machine learning core |
| LSM6DSOX - Step-By-Step Tutorial, Part 1 of 5: Introduction | Machine learning core |
| LSM6DSOX - Step-By-Step Tutorial, Part 2 of 5: Data Collection | Machine learning core |
| LSM6DSOX - Step-By-Step Tutorial, Part 3 of 5: Labeling and Features Extraction | Machine learning core |
| LSM6DSOX - Step-By-Step Tutorial, Part 4 of 5: Device Tree Generation | Machine learning core |
| LSM6DSOX - Step-By-Step Tutorial, Part 5 of 5: Register and Configuration | Machine learning core |
To stay in touch with us:
Use the following opportunities to make sure you stay up to date with the latest information:
Join the MEMS and Sensors Community to learn about new releases, share insights or ask our team of experts your technical questions. Or subscribe to our MEMS and Sensors quarterly newsletter.
