Low-power analog architecture for brain-machine interfaces
An ultra-low-power circuit for wireless neural recording and stimulation is provided. The circuit includes a neural amplifier with adaptive power biasing for use in multi-electrode arrays and a decoding and/or learning architecture. An impedance-modulation telemetry system provides low-power data telemetry. Also, the circuit includes a wireless link for efficient power transfer, and at least one circuit for wireless stimulation of neurons.
1. An ultra-low-power microelectronic system for wireless neural recording capable of being completely implanted in the brain, comprising:
a plurality of neural amplifiers with adaptive power biasing used in multi-electrode arrays, wherein said neural amplifiers comprise a gain stage implemented using an all-subthreshold and folded cascode architecture to allow low-voltage operation and a bandwidth limiting stage to maintain a constant overall bandwidth and to vary a bias current of said gain stage so as to adjust a noise per unit bandwidth of said gain stage;
an adaptive learning architecture for neural signal decoding that uses a continuous-time analog gradient descent algorithm to learn a decoding filter via supervised learning or other learning techniques that is implantable in a brain;
an impedance-modulation telemetry system for low-power data telemetry used by a plurality of brain-machine interfaces to transmit RF data from the brain flexibly and efficiently so as to permit power consumption to occur outside a body rather than within the brain;
a wireless link for efficient power transfer.
2. The system of claim 1 further comprising at least one circuit for wireless stimulation of neurons.
3. The system of claim 1 further comprising an impedance-modulation wireless link in conjunction with a brain-machine interface.
4. The system of claim 3 , wherein the link uses near-field inductive coupling between the implanted system and an external transceiver.
5. The system of claim 4 , wherein the link is designed to minimize power consumption in the brain-implanted system or support high data rates from the implanted to the external system.
6. The system of claim 1 , wherein the analog decoding and learning architecture includes a data compression stage.
7. An ultra-low-power circuit for wireless neural recording and stimulation comprising:
a neural amplifier with adaptive power biasing for use in multi-electrode arrays, wherein said neural amplifier comprises a gain stage implemented using an all-subthreshold and folded cascode architecture to allow low-voltage operation and a bandwidth limiting stage to maintain a constant overall bandwidth and to vary a bias current of said gain stage so as to adjust a noise per unit bandwidth of said gain stage;
an adaptive learning architecture for neural signal decoding that uses a continuous-time analog gradient descent algorithm to learn a decoding filter via supervised learning or other learning techniques that is implantable in a brain;
an impedance-modulation telemetry system for low-power data telemetry used by a plurality of brain-machine interfaces to transmit RF data from the brain flexibly and efficiently so as to permit power consumption to occur outside a body rather than within the brain;
a wireless link for efficient power transfer; and
at least one circuit for wireless stimulation of neurons.
8. The ultra-low-power circuit of claim 7 , wherein said impedance-modulation telemetry system further comprises:
an external unit broadcasting a continuous wave carrier using a class-E amplifier topology; and
an implanted unit that switches its impedance from minimum, a series resonant coil, to a maximum, an open-circuited coil by using a switch in series with a capacitor to amplitude modulate a load at the external unit.
9. The ultra-low-power circuit of claim 7 , wherein said ultra-low-power circuit saves power by entering into a sleep mode during inactivity, and is activated upon receiving a pre-determined signal.
10. The ultra-low-power circuit of claim 7 , wherein said ultra-low-power circuit is implemented as a hybrid analog-digital architecture using DAC-programmable analog circuits that are configurable via an external digital signal processor.
11. The ultra-low-power circuit of claim 7 , wherein said ultra-low-power circuit is usable in conjunction with a neuromotor prosthetic system.
12. An ultra-low-power circuit for wireless neural recording and stimulation comprising:
a neural amplifier with adaptive power biasing for use in multi-electrode arrays, said neural amplifier having a gain stage implemented using an all-subthreshold and folded-cascode architecture to allow low-voltage operation and a bandwidth limiting stage to maintain a constant overall bandwidth and to vary a bias current of said gain stage so as to adjust noise per unit bandwidth of said gain stage;
an adaptive learning architecture for data compression, said adaptive learning architecture implemented using a gradient descent algorithm to learn a decoding filter via supervised learning or other learning techniques that is implantable in a brain;
an impedance-modulation telemetry system for low-power data telemetry, said impedance-modulation telemetry system comprising an external unit broadcasting a continuous wave carrier using a class-E amplifier topology; and an implanted unit that switches its impedance from minimum, a series resonant coil, to a maximum, an open-circuited coil by using a switch in series with a capacitor to amplitude modulate a load at the external unit, said impedance-modulation telemetry system allows a plurality of brain-machine interfaces to transmit RF data from the brain flexibly and efficiently so as to permit power consumption to occur outside a body rather than within the brain;
a wireless link for efficient power transfer; and
at least one circuit for wireless stimulation of neurons.
13. The ultra-low-power circuit of claim 12 , wherein said ultra-low-power circuit saves power by entering into a sleep mode during inactivity, and is activated upon receiving a pre-determined signal.
14. The ultra-low-power circuit of claim 12 , wherein said ultra-low-power circuit is implemented as a hybrid analog-digital architecture using programmable circuits that are configurable via an external processor; including, as one instantiation, a hybrid analog-digital architecture using DAC-programmable analog circuits configurable via an external digital signal processor.
15. The ultra-low-power circuit of claim 12 , wherein said ultra-low-power circuit is usableused in conjunction with a neuromotor prosthetic system.
16. A method of implementing an ultra-low-power circuit for wireless neural recording and stimulation comprising:
providing a micropower neural amplifier with adaptive power biasing for use in multi-electrode arrays, wherein said micropower neural amplifier comprises a gain stage implemented using an all-subthreshold and folded cascode architecture to allow low-voltage operation and a bandwidth limiting stage to maintain a constant overall bandwidth and to vary a bias current of said gain stage so as to adjust a noise per unit bandwidth of said gain stage;
providing an adaptive learning architecture for data compression, said adaptive learning architecture uses a continuous-time analog gradient descent algorithm to learn a decoding filter via supervised learning or other learning techniques that is implantable in a brain;
providing an impedance-modulation telemetry system for low-power data telemetry, said impedance-modulation telemetry system allows a plurality of brain-machine interfaces to transmit RF data from the brain flexibly and efficiently so as to permit power consumption to occur outside a body rather than within the brain;
providing a wireless link for efficient power transfer; and
providing at least one circuit for wireless stimulation of neurons.
17. The method of claim 16 , wherein said impedance-modulation telemetry system further comprises:
an external unit broadcasting a continuous wave carrier using a class-E amplifier topology; and
an implanted unit that switches its impedance from minimum, a series resonant coil, to a maximum, an open-circuited coil by using a switch in series with a capacitor to amplitude modulate a load at the external unit.
18. The method of claim 16 , wherein said ultra-low-power circuit saves power by entering into a sleep mode during inactivity, and is activated upon receiving a pre-determined signal.
19. The method of claim 16 , wherein said ultra-low-power circuit is implemented as a hybrid analog-digital architecture using programmable circuits that are configurable via an external processor; including, as one instantiation, a hybrid analog-digital architecture using DAC-programmable analog circuits configurable via an external digital signal processor.
20. The method of claim 16 , wherein said ultra-low-power circuit is usable in conjunction with a neuromotor prosthetic system.
21. A method of performing the operations of an ultra-low-power circuit for wireless neural recording and stimulation comprising:
performing adaptive power biasing for use in multi-electrode arrays, said neural amplifier having a gain stage implemented using an all-subthreshold and folded-cascode architecture to allow low-voltage operation and a bandwidth limiting stage to maintain a constant overall bandwidth and to vary a bias current of said gain stage so as to adjust a noise per unit bandwidth of said gain stage;
compressing data using a gradient descent algorithm to learn a decoding filter via supervised learning or other learning techniques that is implantable in a brain;
performing low-power data telemetry using a impedance-modulation telemetry system comprising an external unit broadcasting a continuous wave carrier using a class-E amplifier topology; and an implanted unit that switches its impedance from minimum, a series resonant coil, to a maximum, an open-circuited coil by using a switch in series with a capacitor to amplitude modulate a load at the external unit, said impedance-modulation telemetry system allows a plurality of brain-machine interfaces to transmit RF data from the brain flexibly and efficiently so as to permit power consumption to occur outside a body rather than within the brain;
transferring power using wireless link; and
wirelessly stimulating neurons.
22. The method of claim 21 further comprising saving power by entering into a sleep mode during inactivity, and is activated upon receiving a pre-determined signal.
23. The method of claim 21 further comprising providing a hybrid analog-digital architecture using programmable circuits that are configurable via an external processor; including, as one instantiation, a hybrid analog-digital architecture using DAC-programmable analog circuits configurable via an external digital signal processor.
24. The method of claim 21 , wherein said ultra-low-power circuit is used in conjunction with a neuromotor prosthetic system.