Neuronal phase-locked loops

• US 6,581,046 B1
• Filed: 10/09/1998
• Issued: 06/17/2003
• Est. Priority Date: 10/10/1997
• Status: Expired due to Fees

• Current Assignee: Yeda Research and Development Co. Ltd. (Weizmann Institute of Science)
• Original Assignee: Yeda Research and Development Co. Ltd. (Weizmann Institute of Science)
• Inventors: Ahissar, Ehud
• Primary Examiner(s): STARKS, WILBERT L
• Application Number:
  US09/168,672
• Time in Patent Office: 1,712 Days
• Field of Search: 706/15, 331/25, 711/200
• US Class Current: 706/15
• CPC Class Codes: G06N 3/049 Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs

Abstract

A neuronal phase-locked loop (NPLL) that can decode temporally-encoded information and convert it to a rate code is based on an algorithm similar to that of the electronic PLL, but is a stochastic device, implemented by neural networks (real or simulated). The simplest embodiment of the NPLL includes a phase detector (that is, a neuronal-plausible version of an ideal coincidence detector) and a controllable local oscillator that are connected in a negative feedback loop. The phase detector compares the firing times of the local oscillator and the input and provides an output whose firing rate is monotonically related to the time difference. The output rate is fed back to the local oscillator and forces it to phase-lock to the input. Every temporal interval at the input is associated with a specific pair of output rate and time difference values; the higher the output rate the further the local oscillator is driven from its intrinsic frequency. Sequences of input intervals, which, by definition, encode input information, are thus represented by sequences of firing rates at the NPLL'"'"'s output. The NPLL is an adaptive device which can deal with signals whose exact characteristics are not known in advance and can adapt to changing conditions.

Claims

1. A neuronal component, comprising:

an input element for receiving an input signal including a sequence of pulses which are temporally modulated;

an output element for providing an output which is phase locked to said input signal;

at least one first neural network element configured as a phase detector having first and second inputs for receiving temporally modulated signals, and an output for supplying a signal having a firing rate which is a function of a time difference between said temporally modulated signals, said first input being connected to receive said input signal, and said output of said at least one neural network element being connected to said output element; and

at least one second neural network element configured as a rate controlled oscillator having an input connected to said output of said at least one first neural network element, and an output connected to said second input of said at least one first neural network element.

2. A neuronal component according to claim 1, comprising a plurality of said first and second neural network elements connected in parallel.

3. A neuronal component according to claim 2, wherein said first and second neural network elements are connected by diverging and converging pathways.

4. Apparatus for determining at least one of spatial location and movement of an object within a contact area, comprising:

a tactual sensor array having a matrix of individual tactual sensors each detecting spatial information from objects in said contact area;

means for causing said tactual sensor array to scan said contact area at at least one predetermined frequency and in at least one predetermined direction, whereby objects in said contact area trace at least one path across said matrix; and

a processor for processing relative phase and frequency information contained in output signals from sensors in said matrix to determine spatial location and movement of said objects within said contact area;

wherein said processor comprises a phase locked loop circuit comprised of neuronal network elements.

5. Apparatus according to claim 4, wherein said tactual sensor array scans with at least two frequencies, which are one of superimposed and time sequential.

6. Apparatus for determining at least one of spatial location and movement of an object within a contact area, comprising:

a tactual sensor array having a matrix of individual tactual sensors each detecting spatial information from objects in a contact area;

means for causing said tactual sensor array to scan said contact area at at least one predetermined frequency and in at least one predetermined direction, whereby objects in said contact area trace at least one path across said matrix; and

a process for processing relative phase and frequency information contained in output signals from sensors in said matrix to determine spatial location and movement of said objects within said contact area, said processor comprising

at least one phase detector having first and second inputs for receiving temporally encoded signals and an output for supplying a signal having a firing rate which is a function of a time difference of said temporally encoded signals, said first input being connected to receive said output signals from said tactual sensor array, and said output of said phase detector being connected to said output element; and

at least one rate controlled oscillator having an input connected to said output of said phase detector and an output connected to said second input of said phase detector;

wherein at least one of said at least one phase detector and said at least one rate controlled oscillator comprises a neural network component.

7. The signal processing device according to claim 6, wherein said neural network component comprises a neuronal simulator.

8. An object identification apparatus, comprising:

a tactual sensor array having a matrix of individual tactual sensors each detecting spatial information from objects in a contact area;

means for causing said tactual sensor array to scan said contact area at at least one predetermined frequency and in at least one predetermined direction, whereby objects in said contact area view trace at least one path across said matrix; and

a processor for processing relative phase and frequency information contained in output signals from sensors in said matrix to identify an object within said contact area;

wherein said processor comprises a phase locked loop circuit comprised of neuronal network elements.

9. The object identification apparatus according to claim 8, wherein said tactual sensor array scans with at least two frequencies, which are one of superimposed and time sequential.

10. An object identification apparatus, comprising:

a tactual sensor array having a matrix of individual tactual sensors each detecting spatial information from objects in a contact area;

means for causing said tactual sensor array to scan said contact area at at least one predetermined frequency and in at least one predetermined direction, whereby objects in said contact area trace at least one path across said matrix; and

a processor for processing relative phase and frequency information contained in output signals from sensors in said matrix to identify an object within said contact area, said processor comprising

at least one phase detector having first and second inputs for receiving temporally encoded signals and an output for supplying a signal having a firing rate which is a function of a time difference of said temporally encoded signals, said first input being connected to receive said output signals from said tactual sensor array, and said output of said phase detector being connected to said output element; and

at least one rate controlled oscillator having an input connected to said output of said phase detector and an output connected to said second input of said phase detector;

wherein at least one of said at least one phase detector and said at least one rate controlled oscillator comprises a neural network component.

11. A grip control apparatus, comprising:

at least one finger adapted to contact an object within a contact area on a surface of said object;

a tactual sensor array having a matrix of individual tactual sensors disposed on a tip of said finger and contacting said contact area of said object, for detecting spatial information characterizing surface features in said contact area;

means for causing said sensor array to scan said contact area at at least one predetermined frequency and in at least one predetermined direction, whereby surface features in said contact area trace at least one path across said matrix;

a processor for processing relative phase and frequency information contained in output signals from detectors in said matrix to determine movements of said objects; and

means for regulating a force exerted on said contact area in response to an output of said processor;

wherein said processor comprises a phase locked loop circuit comprised on neuronal network elements.

12. A grip control apparatus, comprising:

at least one finger adapted to contact an object within a contact area on a surface of said object;

a tactual sensor array having a matrix of individual tactual sensors disposed on a tip of said finger and contacting said contact area of said object, for detecting spatial information characterizing surface features in said contact area;

means for causing said sensor array to scan said contact area at at least one predetermined frequency and in at least one predetermined direction, whereby surface features in said contact area trace at least one path across said matrix;

a processor for processing relative phase and frequency information contained in output signals from detectors in said matrix to determine movements of said objects; and

means for regulating a force exerted on said contact area in response to an output of said processor;

said processor comprising

at least one phase detector having first and second inputs for receiving temporally encoded signals and an output for supplying a signal having a firing rate which is a function of a time difference of said temporally encoded signals, said first input being connected to receive said output signals from said tactual sensor array, and said output of said phase detector being connected to said output element; and

at least one rate controlled oscillator having an input connected to said output of said phase detector and an output connected to said second input of said phase detector;

wherein at least one of said at least one phase detector and said at least one rate controlled oscillator comprises a neural network component.

13. A neuronal signal processor comprising:

at least one neuronal phase detector comprising a neural component composed of a single layer of at least one neuron, a first and second input to each neuron in the single layer, and an output from each neuron in the single layer having a value that is one of an AND function and an OR function of its inputs, and

at least one neuronal oscillator comprising a neural component being composed of a single layer of at least one neuron having an intrinsic oscillating frequency, a spiking input train to and a spiking output train from each neuron of the single layer of the neural component, where the rate of the spiking input train to each said neural component controls the oscillating frequency of the neuronal oscillator,

said spiking input train to the neuronal oscillator being the output of the neuronal phase detector, and said spiking output train of said neuronal oscillator being one of the inputs to the neuronal phase detector.

14. A neuronal signal processor according to claim 13 wherein all outputs of the single layer of the at least one neuronal phase detector compose a vector the value of which is proportional to the phase difference between the first and second, inputs.

15. A neuronal signal processor according to claim 13 wherein the single layer of the at least one neuronal phase detector is composed of a plurality of neurons, and the first inputs to the plurality of neurons is differentially delayed to at least a portion of the plurality of neurons.

16. A neuronal signal processor according to claim 13 wherein the control of the oscillating frequency of the at least one neuronal oscillator is one of an excitatory and inhibitory control.

17. A neuronal signal processor according to claim 13 wherein the single layer neural of the at least one neuronal oscillator is composed of a plurality of neurons which are inter-connected.

18. A neuronal phase detector comprising a neural component composed of a single layer of at least one neuron, a first and second input to each neuron in the single layer, and an output from each neuron in the single layer having a value that is one of an AND function and an OR function of its inputs.

19. A neuronal phase detector according to claim 18 wherein all outputs of the single layer compose a vector the value of which is proportional to the phase difference between the first and second inputs.

20. A neuronal phase detector according to claim 18 wherein the single layer is composed of a plurality of neurons, and the first inputs to the plurality of neurons is differentially delayed to at least a portion of the plurality of neurons.

21. A neuronal oscillator comprising a neural component being composed of a single layer of at least one neuron having an intrinsic oscillating frequency, a spiking input train to and a spiking output train from each neuron of the single layer of the neural component, where the rate of the input spike train to each said neural component controls the oscillating frequency of the oscillator.

22. A neuronal oscillator according to claim 21 wherein the control is one of an excitatory and inhibitory control.

23. A neural oscillator according to claim 21 wherein the single layer is composed of a plurality of neurons which are interconnected.

24. A spatial encoding device comprising:

a sensor array having a matrix of individual sensors each detecting spatial information from objects in a field of view;

a controller controlling said sensor array to control scanning of said field of view at least at one predetermined frequency and in at least one predetermined direction whereby objects in said field of view trace a path across said matrix; and

a processor for processing relative phase and frequency information contained in output signals from detectors in said matrix to determine spatial location and movement of said objects within said field of view;

wherein said processor includes a phase locked loop circuit comprised of

at least one neuronal phase detector comprising a neural component composed of a single layer of at least one neuron, a first and second input to each neuron in the single layer, and an output from each neuron in the single layer having a value that is one of an AND function and an OR function of its inputs, and

at least one neuronal oscillator comprising a neural component being composed of a single layer of at least one neuron having an intrinsic oscillating frequency, a spiking input train to and a spiking output train from each neuron of the single layer of the neural component, where the rate of the spiking input train to each said neural component controls the oscillating frequency of the neuronal oscillator,

said spiking input train to the neuronal oscillator being the output of the neuronal phase detector, and said spiking output train of said neuronal oscillator being one of the inputs to the neuronal phase detector.

25. A spatial encoding device according to claim 24, wherein said sensor array is a video camera.

26. A spatial encoding device according to claim 25, wherein said video camera scans with at least two frequencies, which are one of superimposed and time sequential.

27. A spatial encoding device according to claim 24, wherein all outputs of the single layer of the at least one neuronal phase detector compose a vector the value of which is proportional to the phase difference between the first and second inputs.

28. A spatial encoding device according to claim 24, wherein the single layer of the at least one neuronal phase detector is composed of a plurality of neurons, and the first inputs to the plurality of neurons is differentially delayed to at least a portion of the plurality of neurons.

29. A spatial encoding device according to claim 24, wherein the control of the oscillating frequency of the at least one neuronal oscillator is one of an excitatory and inhibitory control.

30. A spatial encoding device according to claim 24, wherein the single layer neural of the at least one neuronal oscillator is composed of a plurality of neurons which are inter-connected.

31. A spatial encoding device comprising:

a sensor array having a matrix of individual sensors each detecting spatial information from objects in a field of view;

a controller controlling said sensor array to scan said field of view at least at one predetermined frequency and in at least one predetermined direction, whereby objects in said field of view trace a path across said matrix; and

a processor for processing relative phase and frequency information contained in output signals from detectors in said matrix to determine spatial location and movement of said objects within said field of view, said processor comprising

at least one neuronal phase detector comprising a neural component composed of a single layer of at least one neuron, a first and second input to each neuron in the single layer, and an output from each neuron in the single layer having a value that is one of an AND function and an OR function of its inputs, and

at least one neuronal oscillator comprising a neural component being composed of a single layer of at least one neuron having an intrinsic oscillating frequency, a spiking input train to and a spiking output train from each neuron of the single layer of the neural component, where the rate of the spiking input train to each said neural component controls the oscillating frequency of the neuronal oscillator,

said spiking input train to the neuronal oscillator being the output of the neuronal phase detector, and said spiking output train of said neuronal oscillator being one of the inputs to the neuronal phase detector.

32. A spatial encoding device according to claim 31, wherein all outputs of the single layer of the at least one neuronal phase detector compose a vector the value of which is proportional to the phase difference between the first and second inputs.

33. A spatial encoding device according to claim 31, wherein the single layer of the at least one neuronal phase detector is composed of a plurality of neurons, and the first inputs to the plurality of neurons is differentially delayed to at least a portion of the plurality of neurons.

34. A spatial encoding device according to claim 31, wherein the control of the oscillating frequency of the at least one neuronal oscillator is one of an excitatory and inhibitory control.

35. A spatial encoding device according to claim 31, wherein the single layer neural of the at least one neuronal oscillator is composed of a plurality of neurons which are inter-connected.

36. Apparatus for determining at least one of spatial location and movement of an object within a contact area, comprising:

a tactual sensor array having a matrix of individual tactual sensors each detecting spatial information from objects in said contact area;

a controller controlling said tactual sensor array to scan said contact area at least at one predetermined frequency and in at least one predetermined direction, whereby objects in said contact area trace at least one path across said matrix; and

a processor for processing relative phase and frequency information contained in output signals from sensors in said matrix to determine spatial location and movement of said objects within said contact area;

said processor comprising a phase locked loop circuit comprised of;

at least one neuronal phase detector comprising a neural component composed of a single layer of at least one neuron, a first and second input to each neuron in the single layer, and an output from each neuron in the single layer having a value that is one of an AND function and an OR function of its inputs, and

at least one neuronal oscillator comprising a neural component being composed of a single layer of at least one-neuron having an intrinsic oscillating frequency, a spiking input train to −

and a spiking output train from each neuron of the single layer of the neural component, where the rate of the spiking input train to each said neural component controls the oscillating frequency of the neuronal oscillator,

said spiking input train to the neuronal oscillator being the output of the neuronal phase detector, and said spiking output train of said neuronal oscillator being one of the inputs to the neuronal phase detector.

37. Apparatus according to claim 36 wherein all outputs of the single layer of the at least one neuronal phase detector compose a vector the value of which is proportional to the phase difference between the first and second inputs.

38. Apparatus according to claim 36 wherein the single layer of the at least one neuronal phase detector is composed of a plurality of neurons, and the first inputs to the plurality of neurons is differentially delayed to at least a portion of the plurality of neurons.

39. Apparatus according to claim 36 wherein the control of the oscillating frequency of the at least one neuronal oscillator is one of an excitatory and inhibitory control.

40. Apparatus according to claim 36 wherein the single layer neural of the at least one neuronal oscillator is composed of a plurality of neurons which are inter-connected.

41. Apparatus according to claim 36, wherein said tactual sensor array scans with at least two frequencies, which are one of superimposed and time sequential.

42. An object identification apparatus, comprising:

a tactual sensor array having a matrix of individual tactual sensors each detecting spatial information from objects in a contact area;

a controller for controlling said tactual sensor array to scan said contact area at least at one predetermined frequency and in at least one predetermined direction, whereby objects in said contact area view trace at least one path across said matrix; and

a processor for processing relative phase and frequency information contained in output signals from sensors in said matrix to identify an object within said contact area;

said processor comprising a phase locked loop circuit comprised of at least one neuronal phase detector comprising a neural component composed of

a single layer of at least one neuron, a first and second input to each neuron in the single layer, and an output from each neuron in the single layer having a value that is one of an AND function and an OR function of its inputs, and

at least one neuronal oscillator comprising a neural component being composed of a single layer of at least one neuron having an intrinsic oscillating frequency, a spiking input train to and a spiking output train from each neuron of the single layer of the neural component, where the rate of the spiking input train to each said neural component controls the oscillating frequency of the neuronal oscillator,

said spiking input train to the neuronal oscillator being the output of the neuronal phase detector, and said spiking output train of said neuronal oscillator being one of the inputs to the neuronal phase detector.

43. The object identification apparatus according to claim 42, wherein said tactual sensor array scans with at least two frequencies, which are one of superimposed and time sequential.

44. The object identification apparatus according to claim 42, wherein all outputs of the single layer of the at least one neuronal phase detector compose a vector the value of which is proportional to the phase difference between the first and second inputs.

45. The object identification apparatus according to claim 42, wherein the single layer of the at least one neuronal phase detector is composed of a plurality of neurons, and the first inputs to the plurality of neurons is differentially delayed to at least a portion of the plurality of neurons.

46. The object identification apparatus according to claim 42, wherein the control of the oscillating frequency of the at least one neuronal oscillator is one of an excitatory and inhibitory control.

47. The object identification apparatus according to claim 42, wherein the single layer neural of the at least one neuronal oscillator is composed of a plurality of neurons which are interconnected.

48. A grip control apparatus, comprising:

at least one finger adapted to contact an object within a contact area on a surface of said object;

a tactual sensor array having a matrix of individual tactual sensors disposed on a tip of said finger and contacting said contact area of said object, for detecting spatial information characterizing surface features in said contact area;

a controller for controlling said sensor array to scan said contact area at at least one predetermined frequency and in at least one predetermined direction, whereby surface features in said contact area trace at least one path across said matrix;

a processor for processing relative phase and frequency information contained in output signals from detectors in said matrix to determine movements of said objects; and

a regulator for regulating a force exerted on said contact area in response to an output of said processor;

said processor comprising a phase locked loop circuit comprised of

at least one neuronal phase detector comprising a neural component composed of a single layer of at least one neuron, a first and second input to each neuron in the single layer, and an output from each neuron in the single layer having a value that is one of an AND function and an OR function of its inputs, and

at least one neuronal oscillator comprising a neural component being composed of a single layer of at least one neuron having an intrinsic oscillating frequency, a spiking input train to and a spiking output train from each neuron of the single layer of the neural component, where the rate of the spiking input train to each said neural component controls the oscillating frequency of the neuronal oscillator,

said spiking input train to the neuronal oscillator being the output of the neuronal phase detector, and said spiking output train of said neuronal oscillator being one of the inputs to the neuronal phase detector.

49. A grip control apparatus according to claim 48 wherein all outputs of the single layer of the at least one neuronal phase detector compose a vector the value of which is proportional to the phase difference between the first and second inputs.

50. A grip control apparatus according to claim 48 wherein the single layer of the at least one neuronal phase detector is composed of a plurality of neurons, and the first inputs to the plurality of neurons is differentially delayed to at least a portion of the plurality of neurons.

51. A grip control apparatus according to claim 48 wherein the control of the oscillating frequency of the at least one neuronal oscillator is one of an excitatory and inhibitory control.

52. A grip control apparatus according to claim 48 wherein the single layer neural of the at least one neuronal oscillator is composed of a plurality of neurons which are interconnected.

53. Apparatus for determining at least one of spatial location and movement of an object within a field of view, comprising:

an optical sensor device having a matrix of individual sensors each detecting spatial information from objects in a field of view;

a controller for controlling said optical sensor device to scan said field of view at at least one predetermined frequency and in at least one predetermined direction, whereby objects in said field of view trace a path across said matrix; and

a processor for processing relative phase and frequency information contained in output signals from detectors in said matrix to determine spatial location and movement of said objects within said field of view;

said processor comprising a phase locked loop circuit comprised of

at least one neuronal phase detector comprising a neural component composed of a single layer of at least one neuron, a first and second input to each neuron in the single layer, and an output from each neuron in the single layer having a value that is one of an AND function and an OR function of its inputs, and

at least one neuronal oscillator comprising a neural component being composed of a single layer of at least one neuron having an intrinsic oscillating frequency, a spiking input train to and a spiking output train from each neuron of the single layer of the neural component, where the rate of the spiking input train to each said neural component controls the oscillating frequency of the neuronal oscillator,

said spiking input train to the neuronal oscillator being the output of the neuronal phase detector, and said spiking output train of said neuronal oscillator being one of the inputs to the neuronal phase detector.

54. Apparatus for determining at least one of spatial location and movement of an object within a field of view according to claim 53 wherein all outputs of the single layer of the at least one neuronal phase detector compose a vector the value of which is proportional to the phase difference between the first and second inputs.

55. Apparatus for determining at least one of spatial location and movement of an object within a field of view according to claim 53 wherein the single layer of tie at least one neuronal phase detector is composed of a plurality of neurons, and the first inputs to the plurality of neurons is differentially delayed to at least a portion of the plurality of neurons.

56. Apparatus for determining at least one of spatial location and movement of an object within a field of view according to claim 53 wherein the control of the oscillating frequency of the at least one neuronal oscillator is one of an excitatory and inhibitory control.

57. Apparatus for determining at least one of spatial location and movement of an object within a field of view according to claim 53 wherein the single layer neural of the at least one neuronal oscillator is composed of a plurality of neurons which are interconnected.

58. A slip control apparatus, comprising:

an optical sensor device having a matrix of individual sensors each detecting spatial information from objects in a field of view;

a controller for controlling said optical sensor device to scan said field of view at least at one predetermined frequency and in at least one predetermined direction, whereby objects in said field of view trace a path across said matrix; and

a processor for processing relative phase and frequency information contained in output signals from detectors in said matrix to determine spatial location and movement of said objects within said field of view,

said processor comprises a phase locked loop circuit comprised of

at least one neuronal phase detector comprising a neural component composed of a single layer of at least one neuron, a first and second input to each neuron in the single layer, and an output from each neuron in the single layer having a value that is one of an AND function and an OR function of its inputs, and

at least one neuronal oscillator comprising a neural component being composed of a single layer of at least one neuron having an intrinsic oscillating frequency, a spiking input train to and a spiking output train from each neuron of the single layer of the neural component, where the rate of the spiking input train to each said neural component controls the oscillating frequency of the neuronal oscillator,

said spiking input train to the neuronal oscillator being the output of the neuronal phase detector, and said spiking output train of said neuronal oscillator being one of the inputs to the neuronal phase detector.

59. A slip control apparatus according to claim 58 wherein all outputs of the single layer of the at least one neuronal phase detector compose a vector the value of which is proportional to the phase difference between the first and second inputs.

60. A slip control apparatus according to claim 58 wherein the single layer of the at least one neuronal phase detector is composed of a plurality of neurons, and the first inputs to the plurality of neurons is differentially delayed to at least a portion of the plurality of neurons.

61. A slip control apparatus according to claim 58 wherein the control of the oscillating frequency of the at least one neuronal oscillator is one of an excitatory and inhibitory control.

62. A slip control apparatus according to claim 58 wherein the single layer neural of the at least one neuronal oscillator is composed of a plurality of neurons which are inter-connected.