The emergence of embedded AI has been a recent phenomenon, but it has become a marketing hotspot in the market, playing a key role in the future product market share. Therefore, in industries such as smart homes, smart devices, automotive electronics, and robotics, embedded AI will certainly appear frequently.
Embedded AI, which combines embedded systems with AI, is also known as edge AI. It allows AI algorithms to run on terminal devices without remote execution, resulting in smaller chip sizes, lower costs, and reduced heat generation and power consumption, while also minimizing or eliminating the need to transfer large amounts of data to the cloud, providing a sense of security for usability, speed, privacy, and data security.
How does Beijing Junzheng empower edge AI?
The industrial positioning of MPU determines its mission to deploy edge AI. Beijing Junzheng’s multi-core heterogeneous cross-border processor X2000 belongs to MPU and has a series of MPU technical features. The value of X2000 edge AI includes: faster and closer interaction, as the model executes locally with low latency; more autonomous service capabilities, as it can still provide good service even without a good network; and better privacy protection, as data collection and processing occur locally without the need for uploads.
Empowering edge AI is the most important application positioning of X2000. It showcases the application potential of X2000 from the perspective of AI, as shown in the figure below.
As an edge AI platform, X2000 has comprehensive capabilities: it possesses solid capabilities embedded in the chip, such as the computing power of XBurst 2, instruction set, and memory management, as well as the Beijing Junzheng team’s years of expertise in mastering controllable operating systems and various IoT application development support kits, along with the AI open platform MAGIK. The AI capabilities of X2000, built on its strong capabilities in imaging, video, and connectivity, will undoubtedly make it a rare main control chip in the era of intelligent connectivity.
Combining Software and Hardware to Enhance Fast Face Recognition Capabilities
Taking the low-power, fast-response face recognition solution launched by I Love Solutions Network as an example, we can see the technical route of Beijing Junzheng X2000 in this application.
The 4H99 face recognition solution launched by the partner of I Love Solutions Network integrates computation and recognition, and detection functions, simplifying the process and providing users with a new experience, such as smoother lines and more stable performance. This solution not only supports non-contact face recognition but also supports non-sensing temperature measurement, QR code scanning, RFID, and other functions, widely used in communities, campuses, hospitals, scenic spots, hotels, shopping malls, office buildings, and public service places.
The main control of the 4H99 face recognition solution uses the Junzheng core heterogeneous cross-border processor X2000, which adopts an XBurst®2 logical dual-core + XBurst®0 three-core structure. XBurst® is a 32-bit RISC CPU core with completely independent intellectual property rights from Beijing Junzheng, compatible with the MIPS architecture, and includes 128-bit SIMD extension instructions, hardware floating-point units, and memory management units. The X2000 combines the outstanding computing power of application processors with the real-time control and low power consumption characteristics of microcontrollers.
The algorithms used in 4H99 have received strong support from Junzheng, including optimization of specific algorithms, algorithm porting, DVP scanning proprietary algorithms, and support. In addition to its non-mandatory and non-contact characteristics, 4H99 has several major functional highlights:
1. The solution integrates rich interface functions, and also includes a high-precision face recognition algorithm, combined with a flexible technical architecture, enabling quick recognition within 1 second and providing good face recognition performance, with typical performances including: recognition with hats, recognition with transparent glasses, recognition with makeup, recognition with masks, recognition with beards, and recognition in complex lighting environments.
2. Currently, face recognition access control products based on Android systems are relatively high in cost, and end customers are becoming increasingly strict with cost control. Therefore, the trend of embedded development based on Linux is becoming more evident. In this context, the main control uses the X2000 MIPS architecture face recognition application board 4H99, which has emerged. 4H99 provides SDK source code, allowing customers to delve deeper into logical interaction design.
3. The 4H99 introduces a three-camera interface, which not only supports binocular face recognition but also has a DVP camera interface to provide developers with other richer solutions, such as QR code scanning applications and surveillance.
4. The 4H99 motherboard integrates face recognition interfaces, radio frequency card recognition interfaces, infrared temperature detection interfaces, and scanning module interfaces, allowing for immediate detection of pedestrian surface temperature and alarms, ensuring safety, convenience, and efficiency.
In the field of the Internet of Things, the X2000 is widely used, supporting the implementation of various application projects. The aforementioned face recognition solution is just one of them. The X2000 is an exciting multi-core heterogeneous cross-border MCU. Are you eager to use it to start your smart hardware development journey? Here is a convenient door— the I Love Solutions Network has launched a free trial activity for the Halley5 development board based on Junzheng X2000, which has allowed more developers to gain a deeper understanding of it. Currently, the activity is nearing its end, and friends who want to apply for the development board trial, please scan the QR code below:
