Voice Recognition Systems for Animal Identification

0
702

Recognizing voices is a familiar experience for human beings, whether it is recognizing the voice of a well-known personality on the radio,identifying a friend’s voice over the telephone or hearing the voice of a colleague call out from behind, we are very good at recognizing familiar voices. But what if we had a technology that could identify animals by recognizing their voices?

The innovators are working on voice recognition technology which will work as animal identification (ID) detection system. Imagine how cool it will be if you can understand what your rabbit feeling inside the indoor rabbit hutch.An animal voice pattern recognition algorithm has been developed which will soon be very helpful in identifying different animals. This will not just help the conservation biologists but will also help common people learn new things about the animal kingdom. Here’s how the technology is all set to change our understanding of the animal kingdom.

There are good reasons to develop a non-invasive technology for identification of animals

Understanding of animal behavior and population assessment are two important elements of conservation biology. Field biologists have to face many difficulties while conducting such studies. Some of the challenges faced by them include:

  • Marking and tagging animals is difficult and invasive.
  • Manual data collection and analysis thereafter is a tedious task

The animals have to be caught for marking and tagging, which is not just invasive but also involves lots of risks for both animals and researchers. That’s why the need for a non-invasive technology was felt.

Acoustic analysis based technology – A new frontier in conservation biology

All the species that are capable of producing sounds possess unique acoustic patterns that differ significantly from individual to individual, thus following a common structure typical of the species. Acoustic analysis methodis a technology that identifies individuals or species emitting vocalizations based on these structural patterns.

The technology has been around since 1997, when Acoustic Location Systems were first developed by McGregor at al. But several limitations were faced by the field biologists at that time, like:

  • Devices had to be wireless
  • They had to be waterproof and weatherproof
  • They had to be easily transportable
  • Large memory capacities were required too

Today, the technology has progressed by leaps and bound and it is now possible for today’s devices to have all these features. These devices can localize and track the movements of animals that generate sounds and are now currently called – Automatic Acoustic Survey Systems (AASS).

These smart devices are capable of performing the two main tasks of:

  • Localizing the sound source with precision
  • Identifying the emitter

Till now, AASS have been used mostly to identify marine mammals. The identification systems for terrestrial animals are rather uncommon.

How to these systems work?

Today, the scientists can easily set up networks of miniature recording devices in forests, fields or even on cliff faces to gather large amounts of data, which is important for the working of these identification systems. These systems can automatically detect and recognize animals, birds or insects. With machine learning, statistical analysis, and the enormous data collected by modern devices, the biologists will eventually have a much greater understanding of the animal kingdom, their behavioral patterns, their habitats and even their interactions with each other and humans.

Using advanced machine learning algorithms with the ability to focus on specific sounds while ignoring other noises, the  likelihood of predicting the correct result can be greatly improved.

Field Applications of voice based identification systems

Studying animals in their natural habitats with minimal interference will have many advantages for the researchers and conservationists. Some of the benefits will be:

  • Terrestrial Automatic Acoustic Survey Systems can be used in habitats where visual location is difficult, such as dense tropical forests, reed-beds etc.
  • These systems can be very useful for studying secretive animals and animals which are difficult to observe due to large home ranges or nocturnal activities.
  • They may be even be used to monitor multiple individuals simultaneously, enabling the scientists to study their behavior and interactions between them.
  • Talking of birds, AASS will enable the scientists, ornithologists, conservationists and others to decipher their complex vocal organization and to study their singing interactions. These systems will also help them study the movements of the birds.
  • It can also be used to track large-scale changes in how animals are responding to climate change.

Using Automatic Acoustics Survey Systems, machine learning and data analysis tools the innovators are developing concrete applications with an acoustic platform that will help in estimation of biodiversity of the fauna in different types of environments.

Progress made so far

  • As described earlier, the sound based identification systems have been mainly used in identifying marine animals. Mainly two categories of sounds are studied:
  • Navigational and foraging calls
  • Social and communications calls

By studying these sounds and using pattern recognition, the scientists are able to understand the behavior of marine animals in their natural habitat and their response to climate change.

  • Frogs produce different sounds for different purposes. Males use their voice to attract females and to protect their turfs from enemies. Using advanced voice recognition systems, the researchers from University of Minnesota are trying to identify different types of frogs based on specific sounds. This will also help the scientists to understand their mating behavior and possibly help their dwindling population from going extinct.
  • Acoustic sensors and voice recognition is being used by the scientists to save the declining population of African Elephants. Conservation Metrics, a Microsoft AI for Earth grantee is using machine learning and ‘Elephant Listening Systems’ to monitor Elephant population in Africa and understand their behavior.
  • Today, we have apps that can help even the common people identify birds based on their sounds. For example: Warblr.

There are some challenges to overcome

Despite significant progress in sensor network systems and source localization theory, further development of Automatic Acoustics Survey Systems has been greatly marred by the absence of integrated platforms suitable for monitoring wild animals. Therefore, now the emphasis must be laid on developing fully integrated hardware and software systems that are strong enough tobe deployed in all kind of environments. They should also need to be user-friendly enough to help the biologists with little or no technical expertise.