The development of a sophisticated AI system that could decipher brain waves and convert them into intelligible speech marks a remarkable breakthrough in the technology landscape. The system presents promising prospects for the creation of brain-controlled communication devices.
While the idea of comprehending brain waves and turning them into sensible speech has always remained prominent in the scientific industry, the technology required to turn the idea into reality did not exist until the recent past. With advanced machine learning algorithms, a group of American neuroscientists has succeeded in developing a system that could effectively turn brain waves into speech.
During the development of the system, data from five patients who were undergoing neurosurgery for epilepsy was collected. A variety of electrodes were implanted into the patients’ brains for measuring the electrocorticography activity as they listened to continuous short stories which were being narrated by four different people. Scientists only managed to collect data worth 30 minutes owing to the invasive nature of gathering data during neurosurgery.
The data collected from the patients was used to train a vocoder. Scientists tested the efficiency of the algorithm by testing it to decode voices of people counting from one to nine, data for which had not been included while training the vocoder. As the patients recited digits, their brain waves were recorded and analyzed by deep neural networks. The output from the networks was cleaned and then produced by the vocoder.
The test revealed that the system was able to convert brain waves into intelligible speech with an accuracy of almost 75%. The results were groundbreaking as the amount of data used to train the neural networks was minimal. Researchers believe the results suggest revolutionary implications. However, scientists admitted that refining the system will take almost a decade as data collection for training the algorithm requires implantation of electrodes into the brain.
The team revealed that they will continue refactoring the algorithm to test whether or not it can be used to produce more complex sentences and words from the dataset available currently. The integration of the technology with speech generating devices could be groundbreaking and provide a voice to someone who has lost speech due to injury or accident. In addition to this, the technology can help patients with speech impairments as well. With time, the technology can aid in the proliferation of speech generating devices sales opening new and lucrative opportunities for the manufacturers.