New generation sensor with AI function

Whether 5g or IOT and intelligent driving with constant heat require MEMS sensors to have lower power consumption and higher performance. ST has developed the latest generation of sensors by using the machine learning technology of artificial intelligence, which can improve the overall performance of the system and release the design potential of developers.

Recently, St announced that it has developed a machine learning lsm6dsox sensor. The lsm6dsox sensor integrates a machine learning core, which can classify and process the motion data according to the known motion mode, and replace the first stage task of the main processor in motion tracking. This method can save energy and reduce consumption, speed up fitness recording, health monitoring, personal navigation The running speed of sports applications such as fall detection applications.

MEMS analysis with AI machine learning

The equipment equipped with Italian French semiconductor lsm6dsox can bring users a convenient and responsive “always on” experience, and has little impact on the battery life. Compared with traditional sensors, lsm6dsox adds more FIFO memory space and is equipped with the most advanced high-speed I3C digital interface, which makes the interaction interval between the sensor and the main controller longer, the connection time shorter and the effect of energy saving better.

L the sensor is easy to integrate into mainstream mobile platforms (such as Android and IOS), which can simplify the use process of consumer, medical and industrial intelligent devices. Lsm6dsox includes a 3D MEMS accelerometer and a 3D MEMS gyroscope, and uses a machine learning kernel to track complex motion. It has the lowest power consumption inertial measurement unit on the market (dual-mode high-performance power consumption of 0.55 MA). The sensor hardware runs the machine learning algorithm, saving 10 to 1000 times energy (compared with the application processor method).

L the machine learning kernel works together with the sensor integrated finite state machine logic to perform motion pattern recognition or vibration detection. To create a motion tracking product using lsm6dsox, you can use the open source PC application Weka or cloud neural network to train the machine learning kernel in decision tree classification, generate and set parameters and limits from the sample data, for example, to characterize the acceleration, velocity and magnetic inclination of the detected motion type.

L because it supports free fall detection, wake-up, 6D / 4D direction detection, click and double-click interruption, lsm6dsox can be used in a variety of applications other than motion tracking, such as user interface management and laptop protection. Auxiliary output and configuration options also simplify their optical anti shake (OIS) applications.

Why add artificial intelligence to MEMS sensors?

The artificial intelligence solution can improve the detection ability by 20%, the recognition speed by 50%, and the setting is simple. It only needs 5 steps:

1. The user defines the class to be recognized

2. Collect log data for each class

3. Define the feature that best matches the class to be identified

4. The machine learning tool generates decisions that can be run on lsm6dsox based on these logs and features

5. Configure lsm6dsox and run the program

That’s it!

ST: 20 years of MEMS development history, speak with strength

Feel the 20-year development history of St MEMS from one picture:

MEMS analysis with AI machine learning

MEMS was originally designed for industrial equipment and automotive supplies. With the passage of time, ST has a relatively perfect sensor supply. MEMS is mainly divided into two parts: one is the sensor, which is the acceleration and gyroscope we are familiar with; The other part is the actuator, which is also an important part of MEMS. As an IDM semiconductor company, ST has complete R & D and production. Although the two principles are positive and negative, they can be reused in real design and process.

In 2005, ST’s first 3-axis accelerometer was successfully developed. In 2008, IOS platform award-winning products were widely used by mobile phones, and in 2014

In, it is to deal with a strong market goal: providing friendly support for Android.

ST’s pressure sensor uses its own unique technology to achieve the highest performance and good linearity under high pressure conditions; In the range, the sensitivity is higher. St adopts unique one-time molding packaging technology and market tested packaging technology, with a shipment of more than 300 million pieces.

ST has a wide range of MEMS sensors, including special MEMS sensors in the fields of industry 4.0, high-end industry and intelligent driving. At the same time, there are complete MEMS software solutions, from drivers to software development tools, including underlying drivers, x-cube-mems-xt1, software library and algobuilder tools.

Leave a Reply

Your email address will not be published. Required fields are marked *