What are the challenges in ecosystem surveys?

Using sound as a clue to understand the situation is becoming increasingly important in many areas that support society, from natural environment surveys and animal monitoring to infrastructure conservation. However, these systematic surveys still face many challenges. The first and most prominent example is the shortage of manpower. In addition to the limited number of surveyors with specialized knowledge, surveys require long hours of work in harsh environments such as mountainous regions and areas with high disaster risk, which places a heavy burden on the field and makes it difficult to maintain a continuous survey system.

Furthermore, the costs associated with the survey cannot be ignored. Not only are there labor costs, but the entire process is expensive, including the installation and maintenance of recording equipment and the analysis of huge amounts of sound data.
The conventional method of bringing data back to research institutes for analysis is time-consuming and costly, making it difficult to lead to rapid decision-making.

Additionally, securing communication infrastructure and power sources is a major obstacle at survey sites. In areas where networks cannot reach, such as forests, mountainous regions, remote islands, and remote locations within facilities, real-time data transmission is difficult, and frequent on-site visits are required to replace equipment batteries and for maintenance.
These physical constraints have reduced both the efficiency and accuracy of surveys.

These three issues of "manpower," "cost," and "equipment" are unavoidable bottlenecks when using traditional survey methods. In recent years, edge AI solutions that utilize sound have been attracting attention as a way to solve these three issues.

In this article, we will introduce an edge AI solution that utilizes sound.

Battery-powered voice recognition edge AI system

This system is an edge AI solution that uses the sounds of wild and livestock animals to identify their species on the spot and visualizes the results in real time.
Because it is battery-powered, it can be used in mountainous areas or protected areas where it is difficult to secure power, and can cover a wide area with a minimal configuration.

*As of December 2025, PoC is being created using an evaluation board.

A demo of identifying hawks and eagles

This demo uses Infineon's BLE MCU "CYW20829" and MEMS Mic "IM69D130." The AI model that identifies bird sounds was created using a development environment called "DEEPCRAFT™," also provided by Infineon.
As of December 2025, the hub and visualization server that transmits the inference results to the cloud are still under development, so the inference results are sent to a PC via UART and displayed there.

The graph shows which bird the AI identified for each audio segment over time.

• Pink: Golden Eagle

• Blue: Northern Goshawk (Eagle)

• Gray: Unlabelled

When audio is input, it probabilistically classifies which bird is most likely to be making the sound at that time of day.

0:04~0:20: Eagle call playback → Pink graph (representing the probability of an eagle) appears
0:21~0:36: Hawk call playback → Blue graph (representing the probability of hawk detection) appears

Assumed use case

• Monitoring rare animals in protected areas
• Measures to prevent damage caused by birds and animals in agricultural areas (e.g., wild boars)
• Surveys at universities and research institutes

Summary

In this test, we used an AI model that distinguishes between the calls of hawks and eagles to confirm the effectiveness of an edge AI system that utilizes sound.
By changing the audio data to be identified, it is possible to equip the AI model with the ability to identify a variety of sounds, including those of other birds and wild animals, as well as environmental and equipment sounds.
By preparing AI models suited to different uses and purposes, it is possible to create solutions tailored to each application.

Contact us from here

Please feel free to contact us here for more information about this system, to discuss collaboration, or to consider implementation.