Google has recently announced further investment in AI models designed for whale sound recognition. Their latest multi-species model is capable of identifying the vocalizations of 8 different whale species and can even differentiate between various vocal types within two of those species.
This breakthrough is expected to significantly support marine research, particularly with species like the elusive Bryde’s whales, which are difficult to observe.
The AI model is now publicly accessible, allowing academic researchers to download and apply it in their work.
Google’s whale sound identification efforts have been ongoing since 2018, in collaboration with the National Oceanic and Atmospheric Administration’s (NOAA) Pacific Islands Fisheries Science Center.
Initially focused on detecting humpback whale sounds, the team successfully developed a model that accurately identified these calls, leading to discoveries of whale behavior patterns and even uncovering new habitats.
Additionally, Google partnered with Canada’s Department of Fisheries and Oceans (DFO) and Rainforest Connection to create detection models for Southern resident killer whales, an endangered species.
These models were integrated into an underwater microphone network that reports the whales’ locations to relevant authorities in real time.
Currently, Google’s AI can recognize the sounds of humpback whales, killer whales, blue whales, fin whales, minke whales, Bryde’s whales, North Atlantic right whales, and North Pacific right whales.
Given the wide frequency range of whale sounds—from the low 10 Hz of blue whales to the high 120 kHz of toothed whales—this model represents a pioneering achievement in the field of marine biology.
The technology relies on converting audio into a time-frequency spectrogram, where each time window represents a 5-second sound clip.
These spectrograms are then processed using the Mel scale, a method for adjusting frequencies, which helps highlight key features of the whale sounds.
Compression and regularization techniques are employed to further refine the audio, enabling classification into distinct whale sound categories.
To ensure the accuracy of the model, especially given the various background noises found in the ocean, researchers incorporated a wide range of ambient sounds during training.
As a result, the model is particularly effective in distinguishing the sounds of minke and Bryde’s whales from other noises.
Keep visiting for more such awesome posts, internet tips, and lifestyle tips & remember we cover,
“Everything under the Sun!”
Follow Inspire2rise on Twitter. | Follow Inspire2rise on Facebook. | Follow Inspire2rise on YouTube
Discover more from Inspire2Rise
Subscribe to get the latest posts sent to your email.
