Vocal prosthesis for songbirds

EZEQUIEL M. ARNEODO; TEJASWY PAILLA; WERNER JIANG; VIKASH GILJA; GERT CAUWENBERGHS; TIMOTHY Q GENTNER

Lake Oncone

Simposio; PEW annual meeting; 2016

Pew Charitable Trusts

Brain Machine Interfaces (BMIs) couple neural activity to external artificial effectors. Even for simple actions, this decoding problem is computationally intensive; as the complexity of the behavior increases computations become intractable. As a result, effective neural prostheses for speech and other natural behaviors remain science fiction. We present plan to overcome these limitations by combining recent advances in machine learning, biophysical modelling of vocal production mechanisms, neuromorphic engineering and systems neuroscience. Using birdsong, the preeminent neurobiological model for complex, learned motor behavior, we plan to synthesize song directly from activity in a bird?s brain. We can reduce the computational complexity of the motor-mapping problem. Instead of trying to solve the very difficult task of mapping neural activity to complex muscle movements, we replace the peripheral output organ with a biophysical model whose low-dimensional dynamics captures the whole complexity of the motor output. We will implement machine learning algorithms on a low-power, neuromorphic hardware platform for real-time computation. If successful, our plan will exemplify a new generation of biomimetic systems.