Tuesday, March 16, 2010

A whale of a research project

George Tzanetakis is developing software that assists researchers in understanding orca sounds. Geoff Howe photograph.

By Tom Hawthorn
Special to The Globe and Mail
March 16, 2010

VICTORIA

Two Vancouver Island researchers eavesdrop on orcas.

For more than four decades, Paul Spong and Helena Symonds have recorded whales passing by their OrcaLab research station on Hanson Island, off the northeast coast of Vancouver Island.

They gathered boxes filled with high-quality analog audio cassettes.

The tapes are filled with magnificent whale sounds — pulsed calls, tonal whistles, echolocation clicks.

Less interesting are the untold hours of background noise and hydrophone hissing.

The tapes run for more than 20,000 hours. That’s two years and three months of around-the-clock listening.

It would take nearly seven years to listen to the orca conversations as part of a daily job, not including lunch or coffee breaks.

Then, the sounds would have to be registered, cross-referenced, and annotated. It would be a painstaking task should anyone ever be foolish to attempt it.

Too much data?

Call in a computer scientist.

“It’s a huge amount of audio,” said George Tzanetakis of the University of Victoria. “It’s impossible for a human to process, even if you had an army of volunteers.”

For the past four years, the professor has been transferring the sounds recorded on analog tapes into digital signals. He is about halfway through the batch.

That’s just the first step.

Now that the sounds can be heard online by researchers (and other curious cetacean sympathizers), the next step has been to develop software capable of identifying patterns.

Think of it as the development of a search engine, based on phonetic units, or “small melodic fragments,” instead of proper nouns.

He is also working on filtering unwanted sounds, so the whales’ clicks and squeals are clearer.

The creation of a large file of searchable whale conversations makes more possible the chance we might one day be able to say, “Parlez-vous orca? Sprechen sie orca? Habla orca?”

This aural archive of orca sounds is known as the orchive.

Mr. Tzanetakis, 34, retains his wonder for the noises made by ocean creatures who navigate and communicate by sound.

“There’s something very haunting about it,” he said.

He reminds himself not to read too much into the sounds he hears.

“They may be having quite a mundane conversation. We don’t know.”

He will say this after hearing a gabfest among one resident pod.

“They’re top-of-the-(food)-chain animals. Their only possible enemy is us. They’re quite confident. They can be quite loud in conversation, because they don’t care who’s listening.”

He has become sensitive to the assault of noise which the whales endure from boats and aircraft.

“It must be like living inside a construction zone all your life. That must be quite annoying.”

The professor has a background in music, not marine biology.

He was born in Grenoble, France, where his father was working on his doctorate. His parents, both Greek nationals, were physics students, finding it opportune to be away from their homeland while it fell under control of a military dictatorship. George was born the year after the fall of the junta.

The family returned to Greece, settling in the Heraklion on Crete. The boy played piano, but chaffed at the discipline of practice, preferring instead to improvise. This led to the happy suggestion he join the junior high marching band, for which he was handed a saxophone with the instruction to teach himself. He became a jazz fan.

University studies in computer science, and more formal training in the sax, eventually led him at Princeton University, where he earned doctorate in computer science. His research focuses on music information retrieval.

He has designed software to scan signals to track the beat and determine the genre of the music.

Mr. Tzanetakis (zanna-TACK-iss) also performs with LaSaM, an eclectic ensemble of composers and performers. Their speciality is creative improvisation. (The playful band name is an acronym for Luminosity and Sounds by Adventurous Musicians.)

The work on orca acoustics will also help in culling helpful sounds from the constant hydrophone recordings made by the VENUS and NEPTUNE networks of underwater observatories.

He does not know whether we will ever understand orca language. Imagine being dropped in a land in which you do not know the language, or even the structure of the language, and you cannot ask anyone what anything means, or the context in which it is said. Good luck.

Because orcas are intelligent, their conversation while eating may not be about eating.

He has noted one similarity to human behaviour.

Young orcas use fewer calls than adults, but do so more often.

“Like when children use a word again and again and again,” he said. “They have a smaller vocabulary.”

His help in deciphering the tapes is all the more remarkable for two facts.

The taping began five years before he was born.

Before joining the project, he knew little about orcas.

“I saw ‘Free Willy.’ That was the extent of my knowledge.”

He has since visited the OrcaLab on Hanson Island, but it was at a time of the year when the resident pods were not in the area.

The closet the computer scientist has come to communing with whales, other than listening in on their conversations, came when he joined tourists on a whale-watching tour.

No comments: