CN111401540B - Neuron model construction method and neuron device - Google Patents
Neuron model construction method and neuron device Download PDFInfo
- Publication number
- CN111401540B CN111401540B CN202010156446.9A CN202010156446A CN111401540B CN 111401540 B CN111401540 B CN 111401540B CN 202010156446 A CN202010156446 A CN 202010156446A CN 111401540 B CN111401540 B CN 111401540B
- Authority
- CN
- China
- Prior art keywords
- signal
- potential
- post
- spike
- neuron
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 210000002569 neuron Anatomy 0.000 title claims abstract description 110
- 238000010276 construction Methods 0.000 title abstract description 5
- 230000004913 activation Effects 0.000 claims abstract description 72
- 230000036279 refractory period Effects 0.000 claims abstract description 49
- 239000012528 membrane Substances 0.000 claims description 41
- 230000001242 postsynaptic effect Effects 0.000 claims description 39
- 238000000034 method Methods 0.000 claims description 28
- 230000000284 resting effect Effects 0.000 claims description 22
- 230000001960 triggered effect Effects 0.000 claims description 21
- 230000008569 process Effects 0.000 claims description 19
- 230000008859 change Effects 0.000 claims description 16
- 230000003044 adaptive effect Effects 0.000 claims description 11
- 230000000996 additive effect Effects 0.000 claims description 11
- 238000001514 detection method Methods 0.000 claims description 8
- 238000013528 artificial neural network Methods 0.000 claims description 6
- 230000028161 membrane depolarization Effects 0.000 claims description 6
- 230000001537 neural effect Effects 0.000 claims description 6
- 230000002336 repolarization Effects 0.000 claims description 6
- 210000000225 synapse Anatomy 0.000 claims description 6
- 230000004044 response Effects 0.000 claims description 5
- 230000003213 activating effect Effects 0.000 claims description 4
- 230000003068 static effect Effects 0.000 claims description 2
- 230000002102 hyperpolarization Effects 0.000 abstract description 4
- 238000012545 processing Methods 0.000 abstract description 4
- 230000007274 generation of a signal involved in cell-cell signaling Effects 0.000 abstract 1
- 101100078097 Mus musculus Msmp gene Proteins 0.000 description 12
- 238000010586 diagram Methods 0.000 description 8
- 230000000638 stimulation Effects 0.000 description 3
- 102000004310 Ion Channels Human genes 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000002964 excitative effect Effects 0.000 description 2
- 230000002401 inhibitory effect Effects 0.000 description 2
- 238000011084 recovery Methods 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 239000000654 additive Substances 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 210000000170 cell membrane Anatomy 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000003111 delayed effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000000670 limiting effect Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000002858 neurotransmitter agent Substances 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 230000000946 synaptic effect Effects 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/06—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
- G06N3/063—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using electronic means
- G06N3/065—Analogue means
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/06—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
- G06N3/061—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using biological neurons, e.g. biological neurons connected to an integrated circuit
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Theoretical Computer Science (AREA)
- Neurology (AREA)
- Molecular Biology (AREA)
- Evolutionary Computation (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Microelectronics & Electronic Packaging (AREA)
- Image Analysis (AREA)
- Logic Circuits (AREA)
Abstract
The application discloses a neuron model construction method and a neuron device constructed by the same, wherein a neuron model is constructed by a white noise generator, a refractory period switch signal generator, a post-spike potential signal generator, an activation threshold potential signal generator, a zero-crossing detector, a trigger and a plurality of post-spike potential filters to simulate white noise signals of neurons, emission of spike signals, a refractory period and a hyperpolarization phenomenon, and a complete model of the neurons from signal receiving to signal processing and spike signal emission to output signal generation is provided.
Description
Technical Field
The present invention relates to a simulation technique of neurons, and in particular, to a method for constructing a neuron model and a neuron device constructed by the same.
Background
Neurons are also called nerve cells, and can conduct bioelectric signals through the change of ion concentration inside and outside cell membranes. Electrical signals are transmitted between neurons through synaptic connections and by neurotransmitters. Many neurons are interconnected to form a nerve center, which is responsible for the perception of external information, the control of body reflection, and the higher neural activities such as thinking and memory.
The receiving, processing and outputting of the bioelectric signals by the neurons is a complex process, and the simulation and the modeling of the bioelectric signals are performed, so that the deep understanding of the working mechanism of the nerve cells is facilitated, and the development of a new artificial neural network algorithm and a new hardware circuit is facilitated.
Disclosure of Invention
In view of the above problems, the present application aims to propose a neuron model capable of simulating the conduction of bioelectric signals of neurons.
The application provides a neuron model construction method, wherein input signals of the model are summation of multiple input signals to obtain a total input signal; the output signals of the model are membrane potential signals and post-synaptic potential signals;
emitting a white noise signal with a white noise generator to simulate white noise in neurons;
a refractory period switching signal generator is used for sending out a refractory period switching signal for simulating the refractory period of the neuron;
sending out a post-spike potential signal by using a post-spike potential signal generator to simulate the process of recovering the membrane potential of the neuron from the hyperpolarized potential to the resting potential;
sending out an activation threshold potential signal by using an activation threshold potential signal generator, wherein the activation threshold potential signal is used for simulating and judging whether the neuron can be activated or not;
realizing zero-crossing detection by using a zero-crossing detector to judge whether the current membrane potential of the model exceeds an activation threshold potential; if the current membrane potential exceeds the activation threshold potential signal, the zero-crossing detector sends out a spike signal to simulate the depolarization and repolarization processes of the neuron membrane potential; if the current membrane potential does not exceed the activation threshold potential signal, not sending a spike signal;
simulating responses of synapses of neurons to transmit spike signals of the neurons to other neurons by using a plurality of post-synaptic potential filters;
controlling a trigger by using a peak signal sent by a zero-crossing detection unit; controlling the triggering of the refractory period switch signal generator, the post-spike potential signal generator and the activation threshold potential signal generator by using a trigger;
summing the total input signal with a white noise signal, weighting with a refractory period switch signal and summing with a post-spike potential signal in sequence to obtain a basic potential signal; the basic potential signal and the activation threshold potential signal are subtracted, then a zero-crossing detector judges whether zero crossing occurs or not, and a peak signal is output when zero crossing occurs; the peak signal triggers a refractory period switch signal generator, a post-peak potential signal generator and an activation threshold potential signal generator through a trigger; summing the basic potential signal and the peak signal to obtain a membrane potential signal; the spike signal also generates a post-synaptic potential signal through a post-synaptic potential filter.
Correspondingly, the present application also proposes a neuron model, which includes: a white noise generator, a refractory period switching signal generator, a post-spike potential signal generator, an activation threshold potential signal generator, a zero-crossing detector, a trigger, and a plurality of post-spike potential filters; wherein,
a white noise generator for emitting a white noise signal, thereby simulating white noise in neurons;
the refractory period switch signal generator is used for sending out a refractory period switch signal, so as to simulate the refractory period of neurons;
the post-spike potential signal generator is used for sending out a post-spike potential signal so as to simulate the process that the membrane potential of the neuron is recovered from the hyperpolarization potential to the resting potential;
the activation threshold potential signal generator is used for sending out an activation threshold potential signal, so that the adaptive change and the accumulation effect of the activation threshold potential of neurons are simulated;
the zero-crossing detector is used for realizing zero-crossing detection so as to judge whether the current membrane potential of the model exceeds an activation threshold potential; if the current membrane potential exceeds the activation threshold potential signal, the zero-crossing detector sends out a spike signal to simulate the depolarization and repolarization processes of the neuron membrane potential; if the current membrane potential does not exceed the activation threshold potential signal, not sending a spike signal;
a plurality of post-synaptic potential filters for simulating responses of synapses of neurons to transfer spike signals of the neurons to other neurons;
summing a total input signal obtained by summing multiple paths of input signals of the model with a white noise signal, weighting the total input signal with a refractory period switch signal, and summing with a post-peak potential signal to obtain a basic potential signal; the basic potential signal and the activation threshold potential signal are subtracted, then a zero-crossing detector judges whether zero crossing occurs or not, and a peak signal is output when zero crossing occurs; the peak signal triggers a refractory period switch signal generator, a post-peak potential signal generator and an activation threshold potential signal generator through a trigger; summing the basic potential signal and the peak signal to obtain a membrane potential signal; the spike signal also generates a post-synaptic potential signal through a post-synaptic potential filter.
Preferably, the white noise signal is intrinsic characteristic noise of neurons.
Preferably, after the refractory period switching signal generator is triggered, the generated refractory period switching signal changes from 1 to 0 and then changes from 0 to 1 within a change duration.
Preferably, there may be an adaptive dynamic process of the activation threshold. Due to the influence of the switching dynamic process of the ion channel, when the activation threshold potential generator is triggered, the activation threshold value is increased from a rest threshold value, activated to a higher hyperpolarization threshold value, and then gradually restored to the rest threshold value state. Such activation thresholds have an additive effect. That is, when the activation threshold potential signal has not yet returned to the resting threshold, if triggered again, an increment is generated based on the current value, and then the change is made to the resting threshold.
Preferably, the post-synaptic potential signal has a connection strength; the connection strength is divided into two parts, the first part is the static connection strength, and the second part is the activated connection strength; the activation connection strength is adaptive to changes and has an additive effect. When the post-synaptic potential is triggered by a spike, the activation connection strength will increase and then gradually change toward zero. If the activation connection strength has not been restored to zero, it is again triggered, due to the additive effect, to produce an increment, which then changes back to zero. The sum of the resting bond strength and the activating bond strength constitutes the bond strength of the postsynaptic potential. The increment of the activation connection strength may be a positive value or a negative value. When the increment is positive, the neurons are connected in an excitatory manner. When the increment is negative, inhibitory connection exists among neurons.
Preferably, the plurality of post-synaptic potential filters generate a plurality of paths of post-synaptic potential signals for inputting to other neuron models to form an artificial neural network.
The complete model of the neuron from receiving a signal to processing the signal, emitting a spike signal and generating an output signal is provided through the established neuron model. The model covers white noise signals of neurons, emission of spike signals, refractory period and hyperpolarization phenomena. The activation threshold of the model may be adaptive, and if a neuron spikes for a short period of time, the activation threshold will be higher due to additive effects, making it more difficult for the neuron to be activated. Adaptive changes of the activation threshold potential model fatigue behavior when neurons send spikes. The post-synaptic potential has a connection strength, which models the connection strength of signals as they pass through synapses between neurons. The neuron model can be used for developing an algorithm or a hardware circuit of an artificial neuron and applied to construction of an artificial neural network.
Drawings
FIG. 1 is a diagram showing the process of membrane potential changes when a neuron spikes;
FIG. 2 is a diagram of one embodiment of a neuron model;
FIG. 3 is a waveform diagram of a neuron refractory period switching signal;
FIG. 4 is a waveform diagram of one representation of a spike signal and a post-spike potential signal;
FIG. 5 is a waveform diagram of another representation of a spike signal and a post-spike potential signal;
FIG. 6 is a waveform of a neuron activation threshold potential signal;
fig. 7 is a waveform diagram of the connection strength of post-synaptic potential.
Detailed Description
The present application will be described in detail below with reference to the accompanying drawings.
FIG. 1 shows the change in membrane potential when a neuron spikes. In the absence of a strong external stimulus potential, the neuronal membrane potential remains at resting potential. When the received external stimulation potential exceeds an activation threshold, the ion channel of the neuron membrane is changed, depolarization and repolarization phenomena occur, and a spike signal is generated. Then the membrane potential of the neuron is hyperpolarized and a negative maximum occurs, after which the membrane potential gradually returns to the resting potential. In this process, the neurons become less sensitive to external electrical stimuli, and this time is called the refractory period.
The process of receiving, processing and outputting spikes of external stimulation signals by neurons can be modeled. FIG. 2 depicts a neuronEmbodiment of the model. Firstly, the multi-channel stimulation signals received by the neurons are summed to obtain I (t) which is used as the total input signal of the neurons. The input signal is summed with a white noise signal n (t) to simulate the effect of white noise in neurons on the input signal. A white noise signal n (t) is generated by a white noise generator in a neuron. The signal is then weighted with the refractory period switching signal ear (t) generated by the refractory period switching signal generator. The refractory period switching signal generator is controlled by a trigger. After the switch signal generator of the refractory period is triggered, the switch signal ear (t) of the refractory period changes from 1 to 0 and then gradually recovers to 1, and the change lasts for a time t ear 。
The dashed line in fig. 3 represents a refractory period switching signal ear (t) with a square wave shape, which changes from 1 to 0 after being triggered, and changes for a duration t ear It remains 0 and then becomes 1. At varying duration t ear In the interior, neurons remain in an absolute refractory state to external stimuli. The solid line in fig. 3 shows another embodiment of the refractory period switching signal ear (t), which changes from 1 to 0 after being triggered, and changes for a duration t ear Gradually changing from 0 to 1. When the refractory period switching signal ear (t) is in this state, the process of slowly recovering the neuronal membrane potential from absolute refractory can be simulated. Over a period of time t ear Thereafter, and when there is no new trigger, the refractory period switching signal ear (t) remains at 1.
The signal is weighted by the refractory period switching signal ear (t), and then summed with the post-spike potential signal psmp (t) to obtain the base potential signal b (t). The post-spike potential signal psmp (t) is generated by the post-spike potential signal generator triggered by the trigger. The duration of change of the post-spike potential signal psmp (t) is t psmp The process for outputting the slow recovery of the membrane potential of the neuron from the hyperpolarized potential to the resting potential. Duration of change t psmp After the end and when there is no new trigger, the post-spike potential signal psmp (t) is kept at 0. Duration t of change of refractory period switching signal ear (t) ear And the duration t of the change of the post-spike potential signal psmp (t) psmp May or may not be equal. FIG. 4 shows that at t psmp A waveform diagram of the post-spike potential signal psmp (t) in the time range is a sawtooth wave, which first reaches a negative maximum value and then returns to a resting potential. T in FIG. 5 psmp Another waveform of the post-spike potential signal psmp (t) is shown in the time range, which changes non-linearly from a negative large value to a resting potential.
The activation threshold potential of the neuron is adaptive to changes. The activation threshold potential signal te (t) is issued by the activation threshold potential signal generator. Fig. 6 is a schematic diagram of the change process of the neuron activation threshold potential signal te (t) after being triggered. te (te) rest Is a rest threshold, is a constant, and if the current active threshold potential signal te (t) is the rest threshold and is not triggered, it will always remain at the rest threshold. When the activation threshold potential generator is triggered, the activation threshold te (t) is at the rest threshold te rest On the basis of (2) to produce an increment te add Activation to a higher hyperpolarisation threshold te 0 Then gradually recovering to a rest threshold value state with the recovery time t te . This activation threshold has an additive effect, given that T is the last time the spike was discharged, then there is te 0 ＝te(T)+te add . That is, when the activation threshold potential signal te (t) has not yet returned to the rest threshold, if triggered again, an increment is generated on the basis of the current value, and then the change is made to the rest threshold.
The basic potential signal b (t) needs to be subtracted from the activation threshold potential signal te (t), and then zero-crossing detection is performed to determine whether the current membrane potential exceeds the activation threshold potential. When the membrane potential exceeds the activation threshold, the neuron emits a spike signal sp (t) that simulates the process of depolarization and repolarization of the neuron's membrane potential. And when the zero-crossing condition is not met, no spike signal is emitted. In the embodiment of fig. 4, the spike signal sp (t) is a triangular wave. In the embodiment of fig. 5, the spike signal sp (t) is an impulse function.
The white noise generator, the refractory period switch signal generator, the post-spike potential signal generator, and the activation threshold potential signal generator may constitute a neuron intrinsic signal generator, all controlled by a trigger. The trigger may be integrated in the neuron intrinsic signal generator. The trigger is controlled by a spike sp (t), and is triggered when it receives the spike.
The spike sp (t) generated by the neuron is transmitted to the post-synaptic potential filter. The post-synaptic potential filter is used for simulating the response of the signal to the transmission process of other neurons through synapses and outputting a post-synaptic potential signal psp (t). There may be many post-synaptic potential filters that can be used to simulate the connection of a neuron to other neurons to form a neural network. The filtered post-synaptic potential signal psp (t) may also be multiplexed. Fig. 5 shows a post-synaptic potential signal psp (t) which is delayed with respect to the spike signal sp (t).
The post-synaptic potential signal psp (t) has a junction strength w (t) derived from the resting junction strength w rest And activation of the joining strength w act (t) two parts. Strength of rest connection w rest Is constant, activates the connection strength w act (t) is adaptive and additive. The connection strength of the post-synaptic potential signal is the sum of the resting connection strength and the active connection strength, i.e., w (t) = w rest +w act (t) of (d). Fig. 7 shows a post-synaptic potential connection strength signal w (t). Dotted line is the resting joint strength w rest And is always kept unchanged. Activating the connection strength w when a trigger of a spike signal sp (t) is received act (t) generating an increment w add Then gradually changes to zero, and the required time is t w . If the activation connection strength has not returned to zero, it is again triggered, due to the additive effect, to produce an increment, which then changes back to zero. Strength of rest connection w rest And activated bond strength w act The superposition of the (t) waveforms is the connection strength signal w (t) of the postsynaptic potential. Activating an increment w of joint strength add And may be positive or negative. When the increment of the activation connection strength is positive, the connection between neurons is excitatory. When the increment of the activation connection strength is negative, the connection between neurons is inhibitory.
And summing the base potential signal b (t) and the spike signal sp (t) to obtain a membrane potential signal mp (t) which is used for outputting the membrane potential change state of the neuron. The membrane potential signal mp (t) and the series of post-synaptic potential signals psp (t) together constitute the output signal of the neuron model. To this end, the model of the neuron in this embodiment completes the conversion from the input signal I (t) to the output membrane potential signal mp (t) and a set of post-synaptic potential signals psp (t).
The white noise generator, the refractory period switching signal generator, the post-spike potential signal generator, the activation threshold potential signal generator, the zero-crossing detector, the flip-flop, and the post-spike potential filters, and the operations of summing, weighting, etc., referred to in the present application, may be implemented by corresponding hardware circuits, or may be implemented by programs on devices such as computers, etc.
Unless defined otherwise, all technical and/or scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. The materials, methods, and examples presented in this application are illustrative only and not intended to be limiting.
Although the present invention has been described in conjunction with specific embodiments thereof, it is evident that many alternatives, modifications, and variations will be apparent to those skilled in the art in light of the teachings of this application and yet remain within the scope of this application.
Claims (12)
1. A neuron model building method is disclosed, wherein, the input signal of the model is a total input signal obtained by summing a plurality of input signals; the output signals of the model are membrane potential signals and post-synaptic potential signals;
emitting a white noise signal with a white noise generator to simulate white noise in neurons;
sending out a refractory period switching signal by using a refractory period switching signal generator, and simulating the refractory period of neurons;
sending out a post-spike potential signal by using a post-spike potential signal generator, and simulating the process of restoring the membrane potential of the neuron from the hyperpolarized potential to the resting potential;
an activation threshold potential signal generator is used for sending out an activation threshold potential signal which is used for simulating and judging whether the neuron can be activated;
realizing zero-crossing detection by using a zero-crossing detector to judge whether the current membrane potential of the model exceeds an activation threshold potential; if the current membrane potential exceeds the activation threshold potential signal, the zero-crossing detector sends out a spike signal to simulate the depolarization and repolarization processes of the neuron membrane potential; if the current membrane potential does not exceed the activation threshold potential signal, not sending a spike signal;
simulating responses of synapses of neurons to transmit spike signals of the neurons to other neurons by using a plurality of post-synaptic potential filters;
controlling a trigger by using a peak signal sent by a zero-crossing detection unit; controlling the triggering of the refractory period switch signal generator, the post-spike potential signal generator and the activation threshold potential signal generator by using a trigger;
summing the total input signal with a white noise signal, weighting with a refractory period switch signal and summing with a post-spike potential signal in sequence to obtain a basic potential signal; the basic potential signal and the activation threshold potential signal are subtracted, then a zero-crossing detector judges whether zero crossing occurs or not, and a peak signal is output when zero crossing occurs; the peak signal triggers a refractory period switch signal generator, a post-peak potential signal generator and an activation threshold potential signal generator through a trigger; summing the basic potential signal and the peak signal to obtain a membrane potential signal; the spike signal also generates a post-synaptic potential signal through a post-synaptic potential filter.
2. The neuron model building method according to claim 1, wherein the white noise signal is intrinsic characteristic noise of neurons.
3. A neuron model building method as defined in claim 1, wherein after the refractory period switching signal generator is triggered, the refractory period switching signal is changed from 1 to 0 and then from 0 to 1 within a change duration.
4. A neuron model building method according to claim 1, wherein the activation threshold potential signal is adaptive and has an additive effect, and when it is a resting threshold and is not triggered, it remains at the resting threshold; when triggered, an increment is generated based on the current threshold, the increment is activated to a higher hyperpolarized threshold, and then the increment is gradually changed to a resting threshold.
5. The neuron model building method according to claim 1, wherein the post-synaptic potential signal has a connection strength; the connection strength is divided into two parts, the first part is the static connection strength, and the second part is the activated connection strength; the activation connection strength is adaptive to changes and has an additive effect.
6. A neuron model building method as defined in claim 1, wherein the plurality of post-synaptic potential filters generate a plurality of post-synaptic potential signals for input to other neuron models to form an artificial neural network.
7. A neuronal device, comprising: a white noise generator, a refractory period switching signal generator, a post-spike potential signal generator, an activation threshold potential signal generator, a zero-crossing detector, a trigger, and a plurality of post-spike potential filters; wherein,
the white noise generator is used for emitting a white noise signal, so that the white noise in neurons is simulated;
the refractory period switching signal generator is used for sending out a refractory period switching signal, so that the refractory period of the neuron is simulated;
the post-spike potential signal generator is used for sending out a post-spike potential signal, so as to simulate the process that the membrane potential of the neuron is recovered from the hyperpolarized potential to the resting potential;
the activation threshold potential signal generator is used for sending out an activation threshold potential signal, so that the threshold potential for simulating and judging whether the neuron can be activated is judged;
the zero-crossing detector is used for realizing zero-crossing detection so as to judge whether the current membrane potential of the device exceeds an activation threshold potential; if the current membrane potential exceeds the activation threshold potential signal, the zero-crossing detector sends out a spike signal to simulate the depolarization and repolarization processes of the neuron membrane potential; if the current membrane potential does not exceed the activation threshold potential signal, not sending a spike signal;
the plurality of post-synaptic potential filters are used for simulating responses of synapses of neurons for transmitting spike signals of the neurons to other neurons in a process;
the trigger is used for controlling the triggering of the refractory period switch signal generator, the post-spike potential signal generator and the activation threshold potential signal generator; the trigger is controlled by a peak signal sent by the zero-crossing detection unit;
summing a plurality of input signals of the device to obtain a total input signal, sequentially summing the total input signal with a white noise signal, weighting with a refractory period switch signal and summing with a post-spike potential signal to obtain a basic potential signal; the basic potential signal and the activation threshold potential signal are subtracted, then a zero-crossing detector judges whether zero crossing occurs or not, and a peak signal is output when zero crossing occurs; the peak signal triggers a refractory period switch signal generator, a post-peak potential signal generator and an activation threshold potential signal generator through a trigger; summing the basic potential signal and the peak signal to obtain a membrane potential signal; the spike signal also generates a post-synaptic potential signal through a post-synaptic potential filter.
8. The neuron apparatus of claim 7, wherein the white noise signal is intrinsic characteristic noise of a neuron.
9. A neuronal device according to claim 7, wherein the refractory period switching signal generator is triggered to generate a refractory period switching signal which changes from 1 to 0 and then from 0 to 1 for a duration of change.
10. The neuron apparatus of claim 7, wherein the activation threshold potential signal is adaptive and has an additive effect that, when it is a resting threshold and is not triggered, remains at the resting threshold; when triggered, an increment based on the current threshold is generated, activating to a higher hyperpolarized threshold, and then gradually changing toward a resting threshold.
11. The neuronal device according to claim 7, wherein the post-synaptic potential signal has a connection strength; the connection strength is divided into two parts, the first part is the resting connection strength, and the second part is the activated connection strength; the activation connection strength is adaptive to changes and has an additive effect.
12. The neuron device of claim 7, wherein the plurality of post-synaptic potential filters generate a plurality of post-synaptic potential signals for input to other neuron devices for forming an artificial neural network.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010156446.9A CN111401540B (en) | 2020-03-09 | 2020-03-09 | Neuron model construction method and neuron device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010156446.9A CN111401540B (en) | 2020-03-09 | 2020-03-09 | Neuron model construction method and neuron device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111401540A CN111401540A (en) | 2020-07-10 |
| CN111401540B true CN111401540B (en) | 2023-04-07 |
Family
ID=71430687
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010156446.9A Active CN111401540B (en) | 2020-03-09 | 2020-03-09 | Neuron model construction method and neuron device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111401540B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112541578B (en) * | 2020-12-23 | 2022-10-28 | 中国人民解放军总医院 | Retinal neural network device |
| CN116362189B (en) * | 2023-03-23 | 2024-07-19 | 珠海横琴脑虎半导体有限公司 | Electroencephalogram signal acquisition chip |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101770560B (en) * | 2008-12-31 | 2013-04-24 | 南方医科大学 | Information processing method and device for simulating biological neuron information processing mechanism |
| US9367797B2 (en) * | 2012-02-08 | 2016-06-14 | Jason Frank Hunzinger | Methods and apparatus for spiking neural computation |
| CN103679265A (en) * | 2013-11-21 | 2014-03-26 | 大连海联自动控制有限公司 | MBUN (multi-characteristic bionic unified neuron) model |
| CN106845634B (en) * | 2016-12-28 | 2018-12-14 | 华中科技大学 | A kind of neuron circuit based on memory resistor |
| US10003483B1 (en) * | 2017-05-03 | 2018-06-19 | The United States Of America, As Represented By The Secretary Of The Navy | Biologically inspired methods and systems for automatically determining the modulation types of radio signals using stacked de-noising autoencoders |
| US10505095B2 (en) * | 2018-01-04 | 2019-12-10 | Government Of The United States Of America, As Represented By The Secretary Of Commerce | Neural member, neural network, and neurological memory |
| US11586884B2 (en) * | 2018-02-08 | 2023-02-21 | University Of Massachusetts | Artificial neurons using diffusive memristor |
-
2020
- 2020-03-09 CN CN202010156446.9A patent/CN111401540B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN111401540A (en) | 2020-07-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111401540B (en) | Neuron model construction method and neuron device | |
| CN106845634A (en) | A kind of neuron circuit based on memory resistor | |
| US11429857B2 (en) | Secure voice signature communications system using local and remote neural network devices | |
| Luo et al. | Real-time simulation of passage-of-time encoding in cerebellum using a scalable FPGA-based system | |
| JPS5981755A (en) | Circuit and method of simulating nerve unit | |
| JP2017509952A (en) | Monitoring a neural network with a shadow network | |
| Aghababaiyan et al. | Joint optimization of input spike rate and receiver decision threshold to maximize achievable bit rate of neuro-spike communication channel | |
| KR20160047581A (en) | Methods and apparatus for implementing a breakpoint determination unit in an artificial nervous system | |
| Kositsky et al. | Dynamical dimension of a hybrid neurorobotic system | |
| CN112541578B (en) | Retinal neural network device | |
| Lee et al. | An electronic neuron with input-specific spiking | |
| Johnson et al. | Fault-tolerant learning in spiking astrocyte-neural networks on FPGAs | |
| Jenderny et al. | Wave digital emulation of a light-modulated central pattern generator | |
| Zins et al. | Reproducing Fear Conditioning of Rats with Unmanned Ground Vehicles and Neuromorphic Systems | |
| Masson et al. | Hybrid circuits of interacting computer model and biological neurons | |
| Saggar et al. | System identification for the Hodgkin-Huxley model using artificial neural networks | |
| Northmore et al. | Spike train processing by a silicon neuromorph: The role of sublinear summation in dendrites | |
| Christodoulou et al. | Temporal Pattern Detection and Recognition Using the Temporal Noisy Leaky Integrator Neuron Model with the Postsynaptic Delays Trained Using Hebbian Learning | |
| Argüello et al. | New trends in computational modeling: a neuroid-based retina model | |
| Luo et al. | A scalable FPGA-based cerebellum for passage-of-time representation | |
| Watanabe et al. | System identification of a motor unit pool using a realistic neuromusculoskeletal model | |
| Wei et al. | A novel astrocyte-mediated self-repairing CPG neural network | |
| Deng et al. | The implementation of feedforward network on field programmable gate array | |
| Luo et al. | A real-time silicon cerebellum spiking neural model based on FPGA | |
| Jiang et al. | A Novel Simulation Model for the Motor Unit Innervation Process |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |