At least 5% of the world’s population suffers from some form of speech ailment. While some are curable, many do not have complete cure. However, scientists from Columbia University in the United States may have just got the solution for it.
Their recent research shows that a technology can read brain signals and translate them into speech. Therefore, this will help several people with speech ailment or paralysis to communicate.
A Vocorder to Translate Brain Signals into Speech
Scientific literature shows that our brain develops signals when we speak or when we imagine someone speaking. The brain also develops patterns when we listen to someone speak. However, deciphering these signals was a tough task for Nima Mesgarani and her team. Therefore, they first used simple computer algorithms like spectrograms that help visualize sound frequencies. But, they did not provide satisfactory results.
Further, Mesgarani thought of using vocorder, the technology behind voice assistants like Amazon Echo and Apple Siri. It can process brain signals if trained on recording people’s voices. To do this, Mesgarani took help from Dr Ashesh Mehta, a neurosurgeon at Northwell Health Physicians Partners Neuroscience Institute.
To measure brain signals, epilepsy patients were made to listen voices of people. Further, people recited numbers 0 to 9, and Masgarani and Dr Mehta took note of the brain signals recorded by the vocorder. Comparing both the signals, researchers cleaned the neural network, similar to artificial intelligence mimicking neuron network in a brain.
As a result, researchers got a robot to speak the numbers, which were 75% accurate. Mesgarani and Dr Mehta feel that further research could help scale up the technology into a product. It could be part of an implant that helps patients convert their thought into words.
