When mathematical modelling helps divers communicate

As diving enthusiasts know, you don't venture out alone to discover the depths, but always in pairs. The Guerche brothers, Gabriel (24) and Jonas (29), discovered diving together only a couple of years ago, when they were students at Sorbonne University. After a week of practice in the Mediterranean, on their way back to Paris, they had an idea: why not improving the underwater communication of divers? They only have a limited range of signs at their disposal to know if everything is going well, to say if their equipment is faulty, to indicate a direction, to inform about their state of health, to exchange information about fauna and flora, to communicate during training sessions, etc. The two brothers then imagined the device that they have been perfecting for the past two years after creating the startup 52 Hertz (whose name is inspired by the frequency of a unique whale song).

The latest version of their invention incorporates an innovation co-developed with the support of Inria, and it is with the Valse project team, specialised in control theory for signal processing, that they have found the scientists who accompany them in their developments. "We quickly wanted to put our idea into practice" recalls Jonas Guerche. As I was finishing my studies at the time (a master's degree in innovation management from Sorbonne University and in renewable energy from Mines Paris Tech), and Gabriel was carrying on with them (a degree in physics from Sorbonne University and then a master's degree in nuclear energy from the University of Paris Saclay), we took the plunge and created 52 Hertz with the status of a 'pépite' (student-entrepreneur)."

The rest of the adventure is told by Gabriel Guerche: "Shortly after this creation, I came into contact with an Inria development officer, to whom I wrote directly to present our need to be accompanied by experts in signal processing. We were very pleasantly surprised that a research centre as prestigious as Inria responded to our request! The description of our project was widely circulated within the institute and we were quickly put in contact with Denis Efimov and Rosane Ushirobira, Inria researchers in Lille."

The device developed by 52 Hertz is a pearl of inventiveness, because in water, sound propagates very poorly: we therefore had to find other means of communication. The Guerche brothers had the idea of fitting a microphone and a small receiver to the mouthpiece of the regulator. The sound of the voice is transmitted from the microphone of one diver to the receiver of the other via mini "radio" boxes, and then it travels from the receiver to the diver's ear, via the bone, allowing him to hear!  "However, the invention of 52 Hertz requires a signal processing stage, in this case the voice, explains Rosane Ushirobira, who is interested in the theoretical aspects of this discipline. Indeed, the voice is distorted by the mouthpiece that the diver holds in his mouth. In the first phase of our collaboration, we applied mathematical filtering methods in order to attenuate this disturbance to the voice."

To do this, the scientists had a large number of sound recordings - letters pronounced individually, or several sounds forming phonemes. "With this data and algorithms whose properties we have mastered, we have succeeded in improving the intelligibility of the voice", the researcher summarises.

"However, this is only an intermediate stage, as the filtering must also be effective underwater! Immersion poses other problems," comments Denis Efimov, head of the Waltz team. Underwater, the frequency of sounds decreases with depth and the bubbles and the diver's breathing create new distortions in the sound: the filtering must be able to distinguish, whatever the diving conditions, the useful part of the signal from these disturbances, which have a level and frequencies close to those of the voice."

Another constraint posed by the 52 Hertz device is that the filtering algorithms must be hyper-efficient, in order to perform their filtering tasks in real time, requiring the minimum amount of computing resources and limiting their energy consumption - a sine qua non condition for being used in diving. Fortunately, Valse researchers are already working on algorithms that are effective in situations where signals are highly disturbed and on methods that can be used in embedded applications: for example, they have already successfully applied the fruits of their research to a connected glasses project with the same constraints.

However, the work does not end there. The algorithms still need to be adapted to underwater voice signals, again using recording data provided by 52 Hertz. This work is entrusted to Nathan Capon, a student at Phelma, an engineering school in the Grenoble area, who is doing his second year internship at Inria. "Part of my work consists of implementing the filtering algorithms in a flexible programming language (Matlab, Python) so that they can be embedded without major difficulty. This internship allows me to directly apply theoretical knowledge, the importance of which I realise. I am learning a lot from Denis and Rosane and from my exchanges with the founders of 52 Hertz, discovering the world of the start-up and the research centre," enthuses the young engineer, who had the chance to take part in tests of the device in Brest at the beginning of July, in the company of Jonas and Gabriel.

For the Valse team, this project is an excellent opportunity to apply, in a relatively short time, methods that they have mastered and developed for a long time. "This collaboration with 52 Hertz, apart from being pleasant on a human level, allows us to confront our work with new use cases," confides Rosane Ushirobira. "Our partner expects concrete results from us, so we are not undertaking new research to meet his needs, but with this project we are taking a fresh look at our methods... It is an opportunity, no doubt later on, for new avenues of reflection or innovation," adds Denis Efimov.

What is the next step? The expertise provided by Inria has enabled the Guerche brothers to consolidate a central element of their invention, but they still need to validate and test the whole thing before considering its use and marketing. "Initially, we intend to offer the device to professional divers (technical, scientific, underwater maintenance and armed forces teams, etc.) and, if it proves successful, to market it to the general public. By the end of the year, we expect to have raised significant funds to support our development," says Jonas Guerche.

The collaboration is therefore proving fruitful, both for the Lille team and for the Finistère startup: proof that in innovation, as in scuba diving, the most important thing is to form a good pair!