Stmicroelectronics Nv · EURONEXT:STMPA
Stmicroelectronics Nv — Technical Presentation (2023)
13 slides from Stmicroelectronics Nv's technical presentation, 2023
Slide 1: ISPU: Intelligent sensor processing unit
ISPU: Intelligent sensor processing unit ISPU: What and Why ISPU commercial products ISPU in Industrial: Anomaly Detection With NanoEdge.AI ISPU in Personal Electronics: Unsupervised on device learning in ISPU
Slide 2: ISPU: What and Why
ISPU: What and Why Intelligent Sensor Processing Unit (ISPU): DSP core HIGHLY-SPECIALIZED for Machine Learning and Processing TINY SILICON Unique solution for TinyML with Machine Learning (ML), Binary Neural Network (BNN), and processing capabilities Lowest power consumption IoT node in the market w
Slide 3: AI in the edge: sensors with an Intelligent Sensor Processing Unit (ISPU)
AI in the edge: sensors with an Intelligent Sensor Processing Unit (ISPU) Moving processing from the MCU to the Intelligent Sensor Processing Unit
Slide 4: What’s inside?
What’s inside? DSP for real-time processing and Artificial Intelligence Raw data from MEMS sensor MEMS Sensor: 3 axis accelerometer 3 axis gyroscope ISPU locally process Sensor hub to collect data from data in a few µW additional external sensors (up to 4) ISPU core Only meaningful processed data St
Slide 5: What’s inside?
What’s inside? DSP for real-time processing and Artificial Intelligence Small Area: enhanced 32-bit RISC Harvard Architecture architecture in 8 kilogates RAM based: 40 kB (program + execution) Full Precision: Floating Point Unit Binary Neural Network convolution accelerator: patented by ST Fast inte
Slide 6: ISM330IS, LSM6DSO16IS 6-axis IMU with embedded ISPU
ISM330IS, LSM6DSO16IS 6-axis IMU with embedded ISPU Ultralow power 6-axis IMU for AI in the edge LGA 14L Configurability 2.5x3x0.86 mm • Gyro FS: from ±125 up to ±2000dps (Pin2Pin compatible with all the ST’s 6-axis IMU) • Axel FS: from ±2 g up to ±16 g • ODR up to 6.6 kHz • SPI / I²C digital interf
Slide 7: ISPU commercial products
ISPU commercial products NanoEdge AI Studio for on-device learning solutions LSM6DSO16ISN ISM330ISN ISPU & NEAI Consumer Industrial 2022 ST ISPU toolchain For programming LSM6DSO16IS ISM330IS ✓ CLI ISPU & ✓ IDE ✓ AlgoBuilder Open market For AI: Consumer Industrial ✓ ISPU.AI 2023 How-to instructions,
Slide 8: NanoEdge AI Studio V3.2 Optimized for ISPU
NanoEdge AI Studio V3.2 Optimized for ISPU ON THE PC ON THE ISPU 1 Create the library, ONCE. 2 Use the library, MANY TIMES. Create and embed a self learning engine Standalone PC solution For anomaly detection, the model is (Win/Linux) self-trained at the Edge.
Slide 9: Anomaly detection and AI: one step closer to the signal
Anomaly detection and AI: one step closer to the signal Possibility to divide the tasks between the sensor and the MCU Edge Anomaly Classification Detection ERROR 1 OK ERROR 2 NOK ERROR 3 ERROR 4 Sensor Pump MCU Cloud
Slide 10: NEAI Studio: how to evaluate the ISPU?
NEAI Studio: how to evaluate the ISPU? Products Dil24 Adapter Eval Boards & Kits Software & Package ISM330ISN STEVAL-MKI233KA STM32 NUCLEO: Nano Edge AI Studio • NUCLEO-F401RE • Industrial X-NUCLEO-IKS01A2 X-CUBE-ISPU STEVAL-STWINBX1
Slide 11: X-CUBE-ISPU contents
X-CUBE-ISPU contents Leverage ISPU libraries & source code examples for low power processing CRC Security Man Down Filter Activity Recognition • High Efficiency (Computation Accelerometer (for Wrist) calibration Resources, Power) • High Customization level Wrist Tilt Gyroscope Calibration • Example
Slide 12: AI in the edge: from model training to activity recognition in few steps
AI in the edge: from model training to activity recognition in few steps Unsupervised on device learning in ISPU Smarter and more efficient solutions with ISPU by observing changes in the data and self-adjusting/reconfiguring the sensors’ operating model
Slide 13: Find inspiration to create your personal application
Find inspiration to create your personal application AI in the edge with ultralow power 6-axis IMU for consumer market A completely new level of capabilities and detection accuracy in smart motion sensors with pattern recognition enabled applications : • Consumer health • Gesture recognition • Activ