Embedded Computing Design

Remote edge AI applications are dependent on the size of the embedded hardware device that carry all the computational power and capabilities for processing edge data at the source. Large, embedded AI accelerators face several challenges in space-constrained systems such as being obtrusive and covert. 

The highly compact tactile sensor features multiple 3D magnetometers through Melexis’ Trixaxis technology– protected against magnetic stray fields.

Parkinson’s disease is one of the most common chronic neurodegenerative diseases associated with aging. It affects body movement with primary symptoms such as tremor, rigidity or muscle stiffness, bradykinesia, and postural instability. By the time of diagnosis, Parkinson’s disease (PD) often has reached middle or late stage, leading to more complications.

After the success of Google’s first-generation scalable distributed training and inference system, DistBelief, the Google Brain team, in collaboration with Alphabet, built the second-generation system for the implementation and deployment of large-scale machine learning models, TensorFlow.

After the success of Google’s first-generation scalable distributed training and inference system, DistBelief, the Google Brain team, in collaboration with Alphabet, built TensorFlow, the second-generation system for the implementation and deployment of large-scale machine learning models.

There is an increasing demand for the use of automatic speech recognition in customer service agencies and corporations to reduce procedural complexity. To bring competitive performance and architectural simplicity, an end-to-end automatic speech recognition system has emerged as a savior.

The increasing deployment of embedded AI and ML at the edge has certainly introduced new performance variations from cloud to edge. Despite the abrupt negative change in AI execution performance on the edge device, the adoption of TinyML is a way to move forward.

Within three weeks of the release of Synaptics’ Katana low-power edge AI platform, Lenovo announced its successful partnership with Synaptics to adopt the Katana system-on-chip for their latest devices of the Yoga family of tablets. Taking advantage of the low-power Katana SoC architecture and energy-efficient AI software, Lenovo plans to deliver premium voice communication and voice assistant features in Lenovo’s Yoga 11.

Syntiant introduced a reference design for true wireless earbuds to enable faster time to market for ODMs and OEMs. The ultra-low-power design comes with the NDP100 neural decision processor that provides more than 100x performance and 10x throughput at under 140 microwatts for always-on AI voice processing.

We witnessed many advancements at CES 2022, and nothing is left behind in the embedded health industry. Onera Health launched the Onera Biomedical-Lab-on-Chip, an ultra-low-power biosignal sensor subsystem for wearable devices. This biomedical compact chip is designed for processing multiple biosignals creating a massive opportunity for health devices.

Adafruit, a US-based embedded electronic device manufacturer, known for its popular development boards, external sensors, and supporting hardware aiding the maker community. A while ago, the manufacturer released a 16-pin integrated circuit based on the MCP23017 I2C input/output port expander. The MCP23017 uses two I2C pins that can be shared with external I2C devices which provides 16 general purpose pins in return. 

Neural networks are a popular machine learning model, but they demand higher energy consumption and more complex hardware design. Stochastic computing is an efficient way to balance the trade-off between hardware efficiency and computing performance. However, stochastic computing witnesses low accuracy of ML workloads due to arithmetic units' low data precision and inaccuracy.

Prediction of text readability has become an area of research due to the increasing complexity in the educational content.

When it comes to the operating system required for embedded systems, the Linux distros have been the most favorable choice for developers. The reason being, the framework has made it convenient for developers to choose their target operating system. 

Developing large-scale machine learning solutions has always been a constant struggle.

There was a lot of hype around the original BeagleBoard Foundation’s BeagleV StarFive RISC-V SBC that did not make it to mass production. StarFive knew the need to design a replacement board was inevitable. With many of the features from BeagleV, the RISC-V chipmaker released the VisionFive V1 single board computer featuring the StarFive JH7100 vision SoC. 

All the technological advancements in developing efficient processor cores have led several semiconductor players to design processor IP using open-source instruction set architecture.

Chinese RISC-V chipmaker StarFive has released Dubhe, a 64-bit CPU IP they claim is the world’s highest-performance RISC-V processor core as it clocks at frequencies of up to 2 GHz. Dubhe targets 12 nm TSMC process technology and supports the vector and hypervisor extensions that were recently added to the RISC-V Foundation’s base instruction set architecture. 

San Francisco-based fabless semiconductor company SiFive has launched its second high-performance processor P650 claiming to be the fastest licensable RISC-V processor IP core in the market. Within 6 months of the initial launch of the P550, the manufacturer realized the need for some significant upgrades in terms of the performance directly linking to the clock speed frequency.

With Industry 4.0 kicking in, the need for autonomous robots to perform inference at the edge is increasing exponentially. The integrated sensors on the robotic platform have been an important aspect in the design for robot localization, navigation, and obstacle avoidance.