Building AI-Based Solutions Using PSoC™ 6, AIROC™ Wi-Fi, XENSIV™, SensiML and Picovoice

Machine Learning heavily influences existing and future products to drive innovation in a world aspiring for automation to reduce the reliance on human decisions for everyday tasks. There are already machine learning deployments in everyday products with features like NLU (Natural Language Understanding), vision-based classification (Object/Face), and sensor-based predictive maintenance/anomaly detection. These use cases started with cloud services that generated lots of streaming data resulting in high data cost and latency issues. Because of this, edge-based implementations are becoming more pervasive and in great demand.

At Infineon, we provide a rich set of tools and software that are either Infineon developed or partner-enabled solutions to support three types of developers. The first type of developer already has a model or a dataset and needs to figure out how to deploy it onto a microcontroller and optimize for their chosen system. We provide ModusToolbox™ for Machine Learning to help these developers, which is a set of tools and software that allows developers to take their models and deploy them on our microcontrollers, for example, PSoC™ 6.

The second type of developer is not looking to handcraft their own algorithms and invest significant resources in testing, experimenting, and analyzing every design aspect. They want algorithms to be generated based on data labeling to reduce the time it takes to deploy and develop. They want to capture actual data from real hardware to provide accuracy in classification and deploy it on target hardware for validation and testing. This is where we have partnered with SensiML, which provides auto-generated machine learning inferencing code for complex sensors like motion and vibration pattern recognition. SensiML offers rich development tools, including their AutoML analytics engine. Infineon provides the PSoC™ 6S2 Wi-Fi BT Pioneer Kit and IoT Sense Expansion Kit, which includes a XENSIV™ MEMS microphone and barometric air pressure, Bosch 9-Axis Absolute Orientation sensor, and Piezo Analog Microphone. Combining SensiML and Infineon product offerings, developers will get maximum flexibility that's cost-effective when creating an innovative machine learning product.

The third type of developer is looking for a ready-made solution and only wants to configure the parameters for their implementation. One of these ready-made solutions is a result of Infineon's collaboration with Picovoice. Picovoice provides an accurate and private-based voice platform allowing developers to create their own edge-based voice assistants. Using the Picovoice console, developers can generate their own custom wake word and set of intents as well as attributes to truly make an enriched voice assistant.

Having support for various skill levels and expected deployment flows, Infineon services all types of machine learning developers, from experts to users to beginners. Having this and a combination of our AIROC™ Wi-Fi devices, we can perform digital twin and reporting functionality to the cloud to keep the devices connected but intelligent and secure. Combining all this excites us with the prospect of seeing true innovation unlocking the potential of a great machine learning project.

Want to learn more? Watch the recording of our “Implementing Machine Learning on the Edge with ModusToolbox™, Picovoice and SensiML” webinar below.

Have an idea how to use the technology? Check out our Build AI for IoT design contest for a chance to win hardware and prizes for your project!