A multi-event data parallel transmission method based on nerve conduction mechanism

Abstract

The application discloses a multi-event data parallel transmission method based on a nerve conduction mechanism, which comprises the following steps: inputting data into a neuron LIF model with feedback to obtain a pulse sequence; using different mutually orthogonal wavelet functions to represent the pulse for different events to obtain a wavelet represented pulse sequence; adding the wavelet pulse sequences of all events to form a sending signal and sending it outward; using different wavelets to perform correlation operation on the receiving signal to separate the pulse sequences of each event; and using an integrator to calculate the number of pulses in a time window to restore the numerical data of each event. The application can realize real-time parallel transmission of multi-events, can use a neuron model to represent event information with a pulse sequence, can use mutually orthogonal wavelets as carriers to realize multiplexing, and can realize real-time transmission without waiting in the transmission process, and can be widely applied to real-time transmission of multi-events.

Description

Technical Field

[0001] This invention belongs to the field of communication data transmission technology, specifically relating to a method for parallel transmission of multi-event data based on neural conduction mechanisms. Background Technology

[0002] Modern society is inseparable from the internet, and how to make data transmission on the network faster has always been a hot topic. With the increasing amount and variety of data needing to be transmitted over the network, new demands have been placed on it, such as real-time performance and parallelism. Currently, most networks do not consider the needs for real-time performance and parallelism. When multiple events need to be transmitted, time-division multiplexing is often used. This method leads to data waiting; when previous event data has not been completely sent, newly generated event data must wait, resulting in delayed information received by the receiver. Moreover, as the number of events increases, the waiting time increases, failing to meet real-time requirements. Furthermore, this method is not parallel transmission; the timing of multiple events changes, making it impossible to accurately determine the occurrence time of each event.

[0003] The human nervous system is characterized by its real-time and parallel signal transmission capabilities, making it a highly efficient transmission network. When external stimuli occur, humans can immediately perceive them and even react instantly. The transmission of stimulus signals to the brain via nerves is instantaneous; once a stimulus is generated, the signal is transmitted to the brain in real time. Within the nervous system, various signals are transmitted in parallel, allowing humans to simultaneously perceive different external stimuli, such as light, sound, and temperature, as well as stimuli from different parts of the body. Furthermore, the nervous system is a fast, efficient, and low-energy network, constantly transmitting a large number of signals.

[0004] Given the excellent characteristics of the nervous system, there is a need to design a new data transmission method that can support the real-time parallel transmission of a large number of events based on the mechanism of neural signal transmission, namely, a multi-event data parallel transmission method based on the neural conduction mechanism. Summary of the Invention

[0005] The purpose of this invention is to realize real-time parallel transmission of multiple events. Using a neuron model, event information can be represented by a pulse sequence. Using mutually orthogonal wavelets as carriers enables multiplexing. Furthermore, real-time transmission can be achieved without waiting during the transmission process. This invention can be widely applied to the real-time transmission of multiple events and provides a method for parallel transmission of multiple event data based on neural conduction mechanisms.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a method for parallel transmission of multi-event data based on neural conduction mechanisms, comprising:

[0007] Step 1: Input the data streams describing the events into the corresponding neuron models for pulse encoding to obtain multiple pulse sequences;

[0008] Step 2: Modulate the multiple pulse sequences generated in Step 1 with their respective corresponding wavelets to obtain multiple wavelet pulse sequences;

[0009] Step 3: Superimpose the multiple wavelet pulse sequences from Step 2 and merge them into a single signal, which is then transmitted outwards as the transmission signal;

[0010] Step 4: After receiving the signal from Step 3, the receiver performs correlation operations on the signal and the wavelets corresponding to the multiple descriptive events to recover multiple pulse sequences;

[0011] Step 5: Integrate the multiple pulse sequences recovered in Step 4 to recover the numerical data describing the events.

[0012] Furthermore, the specific steps for pulse encoding of the data stream input corresponding to each event description event in step 1 include:

[0013] Step 11: The data stream describing the event is passed through a linear filter k to extract the signal features of interest;

[0014] Step 12: Input the signal features of interest into a leakage accumulator. When the value V in the accumulator exceeds the set threshold θ, output a pulse sequence.

[0015] Step 13: Pass the output pulse sequence through a feedback filter h and then input it into the accumulator to reduce the value V in the accumulator, so that the accumulator can be triggered again.

[0016] Furthermore, the leakage accumulator satisfies the following expression:

[0017]

[0018]

[0019]

[0020] Among them, I stim The input signal I is generated by the original signal s passing through the linear filter k. sp It is the feedback signal generated by the output pulse r passing through the feedback filter h, where τ is the time constant used to control the rate of change of V. l V is the flip potential, and θ is the threshold value. When V is greater than θ, a pulse will be generated, indicating that there is event information that needs to be transmitted. When V is less than θ, no pulse will be generated, indicating that there is no event information or the event information can be ignored and does not need to be transmitted.

[0021] Furthermore, the frequency of pulses in the pulse sequence is linearly related to the value of the data stream describing the input event; the larger the value of the data stream describing the input event, the denser the pulses generated.

[0022] Furthermore, the neuron model is a LIF neuron model with feedback.

[0023] Furthermore, in step 2, each wavelet is obtained by compressing and translating the wavelet basis function, and all wavelets are pairwise orthogonal. All wavelets need to have the same length. For the compressed wavelet, zero padding is used to make its length the same as the length of the wavelet basis function. At the same time, all wavelets need to have the same energy. For the compressed wavelet, amplitude amplification is used to make its energy the same as the energy of the wavelet basis function.

[0024] Furthermore, in step 3, the wavelets corresponding to each event in the multiple wavelet pulse sequences are mutually orthogonal.

[0025] Furthermore, the correlation operation between the wavelet and the signal corresponding to each descriptive event in step 4 specifically includes:

[0026] Step 41: Locate the starting position of each symbol in the signal and align the wavelet corresponding to the event with the symbol of the signal;

[0027] Step 42: For each symbol of the signal, perform a correlation operation between it and the wavelet corresponding to the event to obtain the correlation value;

[0028] Step 43: Combine the correlation value with the wavelet energy. If the correlation value is greater than the wavelet energy, then... If there exists a pulse that describes the event, then there is no pulse that describes the event.

[0029] Step 44: Repeat steps 41 to 43 until all symbols are traversed in turn to recover the pulse sequence corresponding to the event description.

[0030] Furthermore, the integration of each pulse sequence in step 5 specifically involves: inputting the pulse sequence into the window of the integrator, calculating the number of pulses within the window, and recovering the numerical data of the event described by the pulse sequence.

[0031] Beneficial effects: Compared with the prior art, the present invention has the following advantages:

[0032] 1. Using neuron models to generate pulse sequences representing events helps to extract the characteristics of events, remove noise, and reduce the amount of information that needs to be transmitted;

[0033] 2. In the pulse sequence generated by the neuron model, the frequency of the pulses is proportional to the magnitude of the input data. The receiver can use an integrator to reconstruct the numerical data, making the calculation simple and convenient.

[0034] 3. Wavelet modulation can support parallel transmission of multiple events and achieve multiplexing. The orthogonality of wavelets can be used to easily separate multiple event data.

[0035] 4. Data transmission can be performed without waiting; data can be transmitted as soon as it is generated, achieving real-time data transmission.

[0036] 5. Using wavelets as the carrier, the signal energy is concentrated and the noise resistance is good.

[0037] This invention proposes a method for parallel transmission of multi-event data based on neural conduction mechanisms, which can be used for network data transmission. It utilizes the mechanism of neural conduction to achieve real-time parallel transmission of multiple events. Attached Figure Description

[0038] Figure 1 This is an overall schematic diagram of a multi-event data parallel transmission method based on neural conduction mechanism as described in this invention;

[0039] Figure 2 This is a schematic diagram of the sender in the multi-event data parallel transmission method based on neural conduction mechanism described in this invention;

[0040] Figure 3 This is a schematic diagram of the receiver of a multi-event data parallel transmission method based on neural conduction mechanism as described in this invention;

[0041] Figure 4 This is a schematic diagram of the LIF neuron model with feedback in this invention;

[0042] Figure 5 This is a schematic diagram of the sym8 wavelet basis functions used in this invention. Detailed Implementation

[0043] The invention will now be further explained with reference to the accompanying drawings.

[0044] like Figure 1-3 As shown, this invention provides a method for parallel transmission of multi-event data based on neural conduction mechanisms, comprising:

[0045] Step 1: Input multiple data streams describing events into the corresponding neuron models for pulse encoding to obtain multiple pulse sequences.

[0046] Step 2: Modulate the multiple pulse sequences generated in Step 1 with their respective corresponding wavelets to obtain multiple wavelet pulse sequences.

[0047] Step 3: Superimpose the multiple wavelet pulse sequences from Step 2 and merge them into a single signal, which is then transmitted outwards as the transmission signal.

[0048] Step 4: After receiving the signal from Step 3, the receiver performs correlation operations on the signal and the wavelets corresponding to the multiple descriptive events to recover multiple pulse sequences.

[0049] Step 5: Integrate the multiple pulse sequences recovered in Step 4 to recover the numerical data describing the events.

[0050] In step 1, specifically, using, as Figure 4 The neuron model shown consists of three parts: First, the input signal passes through a linear filter k, which is a 5th-order FIR filter. Considering the simplest case, focusing only on the individual data value itself and ignoring its relationship with preceding and following data, the filter parameters are chosen as {1,0,0,0,0}. Next, a leakage accumulator is input. When the value V in the accumulator exceeds the set threshold θ, a pulse is output. The output pulse sequence passes through a feedback filter h before being input back to the accumulator. Its function is to reduce the value V in the accumulator, allowing it to be triggered again later. The feedback filter h also uses a 5th-order FIR filter, expressed as an exponential function, as follows:

[0051] h(t) = -10 * e -t ,t=1,2,…,5

[0052] The leakage accumulator satisfies the following expression:

[0053]

[0054]

[0055]

[0056] Among them, I stim The input signal I is generated by the original signal s passing through the linear filter k. sp It is the feedback signal generated by the output pulse r passing through the feedback filter h. τ is the time constant, set to 5, used to control the rate of change of V. l The flip potential is set to -0.5. The threshold value θ is set to 1. When V is greater than θ, a pulse will be generated, indicating that there is event information that needs to be transmitted. When V is less than θ, no pulse will be generated, indicating that there is no event information or the event information can be ignored and does not need to be transmitted.

[0057] In step 2, select as follows Figure 5The sym8 wavelet function shown is used as the wavelet basis function. The wavelet basis function is compressed and translated to obtain different orthogonal wavelets. For multiple descriptive events, different wavelets are assigned as their carriers to complete the wavelet modulation of multiple pulse sequences generated in step 1.

[0058] In this embodiment, the wavelet basis function is set to a length of 1024 points, and the wavelet is compressed by 1 / 2, 1 / 4, 1 / 8, and 1 / 16. Then, the wavelet is translated by 32 points to obtain 103 different wavelets. All wavelets are aligned in length to a uniform length of 1024 points, and the amplitude of the wavelets is scaled so that the energy of all wavelets is the same as the energy of the wavelet basis function.

[0059] In step 3, multiplexed transmission occurs: the wavelet pulse sequences generated by each event in step 2 are superimposed and merged into a single signal, which is then transmitted as the transmission signal. Specifically, the multiple signals must be aligned during addition. Since the wavelet length is 1024 points, 1024 points are used as a symbol. For an event, this symbol is either all zeros or a wavelet. The symbols of each event need to be aligned in time. Then, all signals are directly added together to form the transmission signal.

[0060] In step 4, multiplexing occurs: After receiving the signal from step 3, the receiver performs correlation operations on the signal with different wavelets to separate the pulse sequences of each event. Specifically, the wavelet corresponding to each event is used to perform correlation operations with the received signal. Before performing the correlation operation, the starting position of each symbol in each received signal needs to be found, and the wavelet is aligned with the symbols of the received signal. For each symbol of the received signal, a correlation operation is performed between it and the wavelet. If the correlation value is greater than the wavelet energy... This indicates that the event corresponding to the wavelet has a pulse at this time; otherwise, it means that the event has no pulse. Thus, the pulse sequence corresponding to each event can be recovered.

[0061] In step 5, an integrator is used to restore the data: the multiple pulse sequences obtained in step 4 are integrated to restore the numerical data of multiple events. Specifically, due to the characteristics of the neuron model, the frequency of the generated pulses is proportional to the size of the data. Therefore, by using an integrator to calculate the number of pulses that occur within a certain period of time, the numerical data can be restored.

[0062] This invention simulates the signal transmission mechanism of neurons. First, data is passed through a neuron model, outputting a set of pulse sequences. Then, different wavelets are used as pulse carriers for different events, and wavelet modulation is performed. Finally, the wavelet pulse sequences of all events are summed to obtain the transmitted signal. The receiver uses the orthogonality of wavelets to separate the pulse sequences corresponding to each event and uses an integrator to reconstruct the event numerical data. This method has good real-time performance and parallelism, as well as good anti-interference capabilities, and can be widely used for real-time parallel transmission of a large number of events.

[0063] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for parallel transmission of multi-event data based on neural conduction mechanisms, characterized in that, include: Step 1: Input the data streams describing the events into the corresponding neuron models for pulse encoding to obtain multiple pulse sequences; Step 2: Modulate the multiple pulse sequences generated in Step 1 with their respective corresponding wavelets to obtain multiple wavelet pulse sequences; Step 3: Superimpose the multiple wavelet pulse sequences from Step 2 and merge them into a single signal, which is then transmitted outwards as the transmission signal; Step 4: After receiving the signal from Step 3, the receiver performs correlation operations on the signal and the wavelets corresponding to the multiple descriptive events to recover multiple pulse sequences; Step 5: Integrate the multiple pulse sequences recovered in Step 4 to recover the numerical data describing the events.

2. The method for parallel transmission of multi-event data based on neural conduction mechanisms according to claim 1, characterized in that, The specific steps for pulse coding of the data stream describing each event input into the corresponding neuron model in step 1 include: Step 11: The data stream describing the event is passed through a linear filter k to extract the signal features of interest; Step 12: Input the signal features of interest into a leakage accumulator. When the value V in the accumulator exceeds the set threshold θ, output a pulse sequence. Step 13: Pass the output pulse sequence through a feedback filter h and then input it into the accumulator to reduce the value V in the accumulator, so that the accumulator can be triggered again.

3. The method for parallel transmission of multi-event data based on neural conduction mechanisms according to claim 2, characterized in that, The leakage accumulator satisfies the following expression: Among them, I stim The input signal I is generated by the original signal s passing through the linear filter k. sp It is the feedback signal generated by the output pulse r passing through the feedback filter h, where τ is the time constant used to control the rate of change of V. l V is the flip potential, and θ is the threshold value. When V is greater than θ, a pulse will be generated, indicating that there is event information that needs to be transmitted. When V is less than θ, no pulse will be generated, indicating that there is no event information or the event information can be ignored and does not need to be transmitted.

4. The method for parallel transmission of multi-event data based on neural conduction mechanisms according to claim 2, characterized in that, The frequency of pulses in the pulse sequence is linearly related to the value of the data stream describing the input event; the larger the value of the data stream describing the input event, the denser the pulses.

5. The method for parallel transmission of multi-event data based on neural conduction mechanisms according to claim 2, characterized in that, The neuron model is a LIF neuron model with feedback.

6. The method for parallel transmission of multi-event data based on neural conduction mechanisms according to claim 1, characterized in that, In step 2, each wavelet is obtained by compressing and translating the wavelet basis function, and all wavelets are pairwise orthogonal. All wavelets need to have the same length. For the compressed wavelet, zero padding is used to make its length the same as the length of the wavelet basis function. At the same time, all wavelets need to have the same energy. For the compressed wavelet, amplitude amplification is used to make its energy the same as the energy of the wavelet basis function.

7. The method for parallel transmission of multi-event data based on neural conduction mechanisms according to claim 1, characterized in that, Step 3 yields wavelets that are pairwise orthogonal to each event in multiple wavelet pulse sequences.

8. The method for parallel transmission of multi-event data based on neural conduction mechanisms according to claim 1, characterized in that, Step 4, which involves performing correlation operations between the wavelet and the signal corresponding to each descriptive event, specifically includes: Step 41: Locate the starting position of each symbol in the signal and align the wavelet corresponding to the event with the symbol of the signal; Step 42: For each symbol of the signal, perform a correlation operation between it and the wavelet corresponding to the event to obtain the correlation value; Step 43: Combine the correlation value with the wavelet energy. Compare the values; if the correlation value is greater than the wavelet energy... If there exists a pulse that describes the event, then there is no pulse that describes the event. Step 44: Repeat steps 41 to 43 until all symbols are traversed in turn to recover the pulse sequence corresponding to the event description.

9. The method for parallel transmission of multi-event data based on neural conduction mechanisms according to claim 1, characterized in that, In step 5, integrating each pulse sequence specifically involves inputting the pulse sequence into the integrator's window, calculating the number of pulses within the window, and then recovering the numerical data of the event described by the pulse sequence.

Images