Skip to content
Theme
OnDevice Tools

Tech BusinessStart-ups & Funding

Acoustic sensors for farming listen to birds, bees and insects

Chirrup.ai uses microphones and machine learning to measure farmland biodiversity, while AgriSound identifies pollinating bees by their wingbeats. EE Times reports that both firms have moved from trials to commercial deployments with farmers and retailers.

Illustration of a solar-powered farm sensor recording birdsong and bee wingbeats in a field

Key takeaways

  • Chirrup.ai says its acoustic recorders have measured biodiversity on more than 300 farms in eight European countries.
  • AgriSound's solar-powered Polly sensor identifies honeybees and bumblebees by wingbeat frequency and sends data to the cloud about once an hour.
  • Chirrup.ai analyses recordings off site to save battery and cost, while AgriSound processes audio on the device to give growers fast answers.
  • Both companies pair machine learning with review by ornithologists, ecologists or entomologists to keep species identification reliable.
In this article
  1. What the sound sensors do
  2. How Chirrup.ai records birdsong
  3. AgriSound listens for wingbeats
  4. Speed versus battery life
  5. What it means for growers

What the sound sensors do

Acoustic sensors for farming are moving out of the lab and onto working farmland, where microphones and machine learning are being used to track wildlife and pollinating insects. According to EE Times, the British start-up AgriSound has deployed its Polly sensor for the retailer Marks & Spencer and on fruit and sugar beet farms. Chirrup.ai, meanwhile, says its recorders have gathered biodiversity data from more than 300 farms in eight European countries. The EE Times report, published on 16 November 2026, sets out how both companies turn recorded sound into information for growers and land managers.

The approach covers a wide range of noises. EE Times lists birdsong, the wingbeats of pollinating insects and the vibrations made by insects feeding inside plants as signals that can be captured and interpreted. In each case a small recorder gathers evidence about conditions on the land over days or weeks, rather than in a single visit.

How Chirrup.ai records birdsong

Chirrup.ai was launched in March 2022 by founder and chief executive Conrad Young, after he spent the Covid-19 lockdowns thinking about how birdsong could stand in for wider biodiversity. He told EE Times that birdsong is "scientifically valid and benchmarkable". The company's first task was to build a baseline of bird activity, which gives a reference point so later recordings can show whether diversity is rising, falling or holding steady.

Early trials used audiomoths, which are low-cost acoustic loggers. Chirrup.ai has since developed its own smart sensor. It pairs a MEMS analogue microphone, a very small microphone built using micro-electro-mechanical systems, with a 24-bit audio codec. A 240-MHz processor handles the audio and links over Bluetooth, a 4G LTE modem moves data, and a multi-constellation GNSS receiver adds position and time stamps using GPS, GLONASS and Galileo. A rechargeable lithium-polymer battery supplies power.

In the field, several recorders are spaced at least 200 metres apart along the boundary between woodland and open land where birds are present. They are kept clear of roosting sites, paths and roads. Each device listens for two to three weeks before it is collected and returned to Chirrup.ai, where the audio is turned into spectrograms. A spectrogram is a picture of sound that shows which frequencies appear over time, which makes species easier to tell apart.

Machine-learning models then work through those images. Chirrup.ai has trained algorithms to pick up small differences in recordings, such as one species copying another, and to remove background noise. It also uses BirdNET, a machine-learning platform trained on bird species from around the world.

Because bird calls vary so much, people are still central to the process. Technology director Paul Nebel said ornithologists and ecologists have reviewed thousands of hours of audio, comparing their judgements with the model's answers and setting confidence thresholds. He said the checks are needed to be sure the machine learning "doesn't drift off". From its species data, together with hedgerow health, land management, satellite habitat data and each species' place in the food chain, the firm calculates a biodiversity index.

AgriSound listens for wingbeats

AgriSound was founded in 2020 by chief executive Casey Woodward and receives government funding. Its solar-powered sensor, named Polly, listens for honeybees and bumblebees in glasshouses, polytunnels and open fields, and identifies them by the frequency of their wingbeats.

The device combines a MEMS microphone and an acoustic sensor array with sensors for temperature, humidity and ambient light. An onboard processor runs machine-learning code that measures the wingbeat frequency in hertz and separates bee activity from other noise. Results travel to the cloud over 4G, usually once an hour, and a Bluetooth link allows firmware updates.

That software was trained on a library built from thousands of hours of recordings made under controlled conditions, where individual bees were encouraged to fly towards a flower and collect pollen and nectar. Woodward said bees are "very smart and difficult to incentivize".

Marks & Spencer has used the sensors to gauge pollinator numbers, and EE Times reports that the devices can also be found on strawberry and sugar beet farms and other sites across the U.K. AgriSound's technology has reached almond orchards in Spain and Portugal, blueberry crops in the United States and avocado farms. Woodward said the company's glasshouse work has helped growers cut the share of misshapen fruit.

Speed versus battery life

Illustration: battery, AI, microphone, bluetooth and the labels Acoustic Sensors, Agrisound, Biodiversity

The two companies have made opposite trade-offs. Chirrup.ai sends recordings away for analysis, which keeps the devices cheaper and their batteries longer-lasting. Nebel said processing audio on the device itself would require "a bigger battery and probably a solar cell", which would push the cost up.

AgriSound analyses sound on the device because its customers need answers quickly. Woodward said an almond flowering season may last only 10 days, so growers want recommended interventions to arrive fast. Chirrup.ai is answering a slower question. Young said farmers want to understand how nature is changing year on year, so the company provides what he called "an annual snapshot".

What it means for growers

For farmers, the appeal is regular monitoring of things that are hard to count by hand. Pollinator activity feeds directly into cropping, since poor pollination shows up as misshapen fruit, and AgriSound says its glasshouse work has reduced that problem. Bird populations are also widely treated as a signal of the health of the wider environment, which is why Chirrup.ai's index is aimed at land management as much as at wildlife records.

Both firms are widening what they listen for. Chirrup.ai is working to add weather data so seasonal changes can be judged in context, and Young said the company could extend bioacoustic sensing to insects, mammals and even livestock. EE Times describes AgriSound's move into livestock as very recent.

The deployment figures deserve some caution. They come from the two companies in a single EE Times report, with no independent verification included. The report also makes clear that accuracy still rests on people, since both firms rely on expert review to keep their models aligned with what a trained listener would hear.

Sources

  1. The Sound of Smarter FarmingEE Times

Tags

Spotted a mistake? See our corrections policy. Read how News by OnDevice works in our editorial policy.