Case Studies

Background

In 2020, Azure planned to build fast-track platform for developers working on AI applications at the edge. They reached out to us to develop the Audio System on Module as part of their suite, requiring low-latency processing and robust audio input handling.

Our Solution

We offered the Azure Team a tailored Microphone Array Solution, featuring hardware design, embedded software, and reference low-level code on host designs.

Implementation

We collaborated closely with Azure team to identify the most efficient approach their spec, reviewing every hardware iteration before production. A key aspect of our implementation is developing the low-level USB code for customized duplex communication between the host and the board through USB. This brings convenience to users, allowing them to control and observe the board without needing to code on the embedded system.

Background

In 2021, Alibaba Cloud launched the Smart Classroom Project, a public welfare initiative aimed at equipping remote classrooms with advanced smart technology. They entrusted us with developing the critical audio component to ensure clear communication within these classrooms.

Our Solution

We engineered a robust audio system featuring cascaded ceiling microphones and ADI's A2B technology for seamless data transmission. Our solution includes a flexible PCBA design on the XMOS hardware platform, enabling real-time status tracking.

Implementation

The whole idea is about using multiple nodes to capture the voice and status data, then send them to the terminal. We successfully ported ADI's A2B protocol stack to the XMOS platform without direct technical support from ADI.

Our scalable design supports over 30 nodes with multi-host expansion, includes a sophisticated node status monitoring system for real-time management and troubleshooting.

Background

For many years, we have partnered with Muyuan, the world's largest pig farming company to develop the next generation of AI livestock health monitoring systems. This system leverages various smart devices, including sensors and audio equipment, to monitor pigs' health and reduce losses.

Our Solution

We utilize advanced machine learning techniques and DSP algorithms to process and classify audio data of pig sounds, allowing for early detection of pigs' diseases and timely responses.

Implementation

Our monitoring solution is a sophisticated system that encompasses audio data collection and processing, training and deploying edge computing models, as well as providing solutions for system upgrades and troubleshooting.

We employ optimization techniques like quantization and pruning to reduce the model size, ensuring it meets the speed and memory requirements for real-time operation on embedded processors.