WO2023101201A1 - 스파이킹 신경망 구조의 픽셀화된 모놀리식 광혈류측정 센서 및 그의 구동 방법 - Google Patents
스파이킹 신경망 구조의 픽셀화된 모놀리식 광혈류측정 센서 및 그의 구동 방법 Download PDFInfo
- Publication number
- WO2023101201A1 WO2023101201A1 PCT/KR2022/015490 KR2022015490W WO2023101201A1 WO 2023101201 A1 WO2023101201 A1 WO 2023101201A1 KR 2022015490 W KR2022015490 W KR 2022015490W WO 2023101201 A1 WO2023101201 A1 WO 2023101201A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- output
- signal
- spike
- neurons
- input
- 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.)
- Ceased
Links
Images
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/02—Detecting, measuring or recording for evaluating the cardiovascular system, e.g. pulse, heart rate, blood pressure or blood flow
- A61B5/024—Measuring pulse rate or heart rate
- A61B5/02416—Measuring pulse rate or heart rate using photoplethysmograph signals, e.g. generated by infrared radiation
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N25/00—Circuitry of solid-state image sensors [SSIS]; Control thereof
- H04N25/10—Circuitry of solid-state image sensors [SSIS]; Control thereof for transforming different wavelengths into image signals
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/02—Detecting, measuring or recording for evaluating the cardiovascular system, e.g. pulse, heart rate, blood pressure or blood flow
- A61B5/024—Measuring pulse rate or heart rate
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/02—Detecting, measuring or recording for evaluating the cardiovascular system, e.g. pulse, heart rate, blood pressure or blood flow
- A61B5/024—Measuring pulse rate or heart rate
- A61B5/02416—Measuring pulse rate or heart rate using photoplethysmograph signals, e.g. generated by infrared radiation
- A61B5/02427—Details of sensor
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/145—Measuring characteristics of blood in vivo, e.g. gas concentration or pH-value ; Measuring characteristics of body fluids or tissues, e.g. interstitial fluid or cerebral tissue
- A61B5/1455—Measuring characteristics of blood in vivo, e.g. gas concentration or pH-value ; Measuring characteristics of body fluids or tissues, e.g. interstitial fluid or cerebral tissue using optical sensors, e.g. spectral photometrical oximeters
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/145—Measuring characteristics of blood in vivo, e.g. gas concentration or pH-value ; Measuring characteristics of body fluids or tissues, e.g. interstitial fluid or cerebral tissue
- A61B5/1455—Measuring characteristics of blood in vivo, e.g. gas concentration or pH-value ; Measuring characteristics of body fluids or tissues, e.g. interstitial fluid or cerebral tissue using optical sensors, e.g. spectral photometrical oximeters
- A61B5/14551—Measuring characteristics of blood in vivo, e.g. gas concentration or pH-value ; Measuring characteristics of body fluids or tissues, e.g. interstitial fluid or cerebral tissue using optical sensors, e.g. spectral photometrical oximeters for measuring blood gases
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7203—Signal processing specially adapted for physiological signals or for diagnostic purposes for noise prevention, reduction or removal
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7225—Details of analogue processing, e.g. isolation amplifier, gain or sensitivity adjustment, filtering, baseline or drift compensation
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N25/00—Circuitry of solid-state image sensors [SSIS]; Control thereof
- H04N25/70—SSIS architectures; Circuits associated therewith
- H04N25/76—Addressed sensors, e.g. MOS or CMOS sensors
- H04N25/77—Pixel circuitry, e.g. memories, A/D converters, pixel amplifiers, shared circuits or shared components
Definitions
- the present disclosure relates to a pixelated monolithic photoplethysmography sensor having a spiking neural network structure and a driving method thereof.
- Photoplethysmography is a sensing technique most commonly used in wearable devices for heart rate monitoring or oxygen saturation measurement. Since the main principle is to measure the change in light signal according to the change in blood flow, the PPG sensor uses a photodiode (PD) that converts the light signal into a current signal and a high-performance analog front end that converts the current signal into digital data. -end, AFE). Recently, a monolithic CMOS PPG sensor that implements both PD and AFE on a single die with a semiconductor integrated circuit process has emerged, showing excellent performance in various aspects such as area used, number of parts, and signal-to-noise ratio performance.
- PD photodiode
- AFE high-performance analog front end
- CMOS image sensor technology Existing monolithic PPG sensors use a distributed LDC (Light-to-Digital Converter) structure or use a plurality of PDs arranged in an array to use pin diodes provided by CMOS image sensor technology. Since the PPG signal is a change in the total sum of light applied to all PDs, in general, in order to maximize the signal-to-noise ratio of the PPG signal, multiple PD outputs are summed with an analog circuit or multiple LDC outputs are summed with a digital adder. method to generate the sensor output.
- LDC Light-to-Digital Converter
- Each PD in the PD array receives a different light input according to its position. For example, a PD closer to the light source receives strong light input, and a PD farther receives weak light input. That is, if each PD output is independently processed, spatial information of the PPG signal can be obtained. When spatial information is obtained, noise due to ambient light can be separated from a signal using signal processing techniques such as spatial filtering. This has not been possible with conventional PPG sensors.
- the output of each PD can be individually extracted from the PD array.
- the CMOS image sensor time-divides the output of each PD pixel and transmits it to the AFE, this method is not suitable for PPG sensors that require real-time output.
- the present invention is to solve the problems of the prior art, and is a pixelated monolithic optical blood flow of a spiking neural network structure capable of outputting a spike signal according to spatial information of each pixelated light measuring neuron (LSN). It aims to provide a measurement (PPG) sensor and a method for driving the sensor.
- PPG measurement
- the present invention is to solve the above-mentioned problems of the prior art, a spiking neural network structure capable of merging a plurality of outputs obtained corresponding to the column direction or row direction of the pixel structure into a single output without a separate adder. It is an object to provide a monolithic photoplethysmography (PPG) sensor and a method for driving the sensor.
- PPG photoplethysmography
- the present invention is to solve the above-mentioned problems of the prior art, and is a pixelated monolith of a spiking neural network structure that can obtain an output from which noise due to ambient light is removed through a simple matrix operation without a conventional complex signal processing technique.
- An object of the present invention is to provide a photoplethysmography (PPG) sensor and a method for driving the sensor.
- PPG photoplethysmography
- each pixelated monolithic photoplethysmography sensor of a spiking neural network structure generates a spike signal, and at least one of a row direction and a column direction a plurality of input neurons arranged in one direction; a plurality of output neurons summing the spike signals applied from the plurality of input neurons in the column direction or the row direction, and generating an output spike signal when the result of the sum is equal to or greater than a predetermined reference value; and a synchronizer converting each of the output spike signals into a digital signal.
- each of the plurality of input neurons may include an optical-to-digital converter that generates the spike signal based on the magnitude of the received light; a first transistor outputting the spike signal to a column input node corresponding to the column direction; and a second transistor outputting the spike signal to a row input node corresponding to the row direction.
- the optical-to-digital converter may generate the spike signal having an intensity proportional to the magnitude of the received light.
- the plurality of input neurons are arranged to form an array including n x m pixels, and the column input node includes the spikes generated from m pixels corresponding to each column of the array. It may include n nodes to which signals are applied, and the row input node may include m nodes to which the spike signals generated from n pixels corresponding to each row of the array are applied.
- the plurality of output neurons include n output neurons each corresponding to any one of the columns constituting the array and m output neurons each corresponding to any one of the rows constituting the array can include
- the synchronizer may convert the output spike signal into a 1-bit digital signal synchronized with a clock signal.
- relative spatial information of light applied to the plurality of input neurons to the photoplethysmography sensor may be derived based on information on the magnitude of each of the digital signals.
- a driving method of a pixelated monolithic photoplethysmography sensor having a spiking neural network structure is based on the magnitude of light received by a plurality of input neurons. generating a spike signal; summing the spike signals applied from the plurality of input neurons in a column direction or a row direction; generating an output spike signal by a plurality of output neurons when the result of the summation is equal to or greater than a preset reference value; and converting each of the output spike signals into a digital signal.
- the generating of the spike signal may include generating the spike signal having an intensity proportional to the magnitude of light received by an optical-to-digital converter, and inputting the spike signal to a column corresponding to the column direction. node and a row input node corresponding to the row direction.
- the converting into a digital signal may include converting the output spike signal into a 1-bit digital signal synchronized with a clock signal.
- PPG pixelated monolithic photoplethysmography
- LSN pixelated light measuring neuron
- a plurality of outputs obtained in correspondence to the column direction or row direction of the pixel structure through the pixelated monolithic photoplethysmography (PPG) sensor of the spiking neural network structure are separated by an adder can be merged into a single output without
- FIG. 1 is a schematic configuration diagram of a pixelated monolithic photoplethysmography sensor having a spiking neural network structure according to an embodiment of the present disclosure.
- FIG. 2 is a diagram showing a schematic circuit diagram of an input neuron according to an embodiment of the present disclosure.
- FIG. 3 is a diagram showing a circuit diagram of an optical-to-digital converter according to an embodiment of the present disclosure.
- FIG. 4 is a diagram illustrating a circuit diagram of an output neuron according to an exemplary embodiment of the present disclosure.
- FIG. 5 is a diagram showing a circuit diagram of a synchronizer according to an embodiment of the present invention.
- FIG. 6 is a graph exemplarily showing photoplethysmography sensor data and noise obtained using a pixelated monolithic photoplethysmography sensor according to an embodiment of the present disclosure.
- FIG. 7 is a graph exemplarily illustrating a result of removing noise caused by ambient light using spatial information using photoplethysmography sensor data according to an embodiment of the present disclosure.
- FIG. 8 is an operation flowchart of a method of driving a pixelated monolithic photoplethysmography sensor having a spiking neural network structure according to an embodiment of the present disclosure.
- the present disclosure relates to a pixelated monolithic photoplethysmography sensor having a spiking neural network structure and a driving method thereof.
- the pixelated monolithic photoplethysmography (PPG) sensor 10 according to an exemplary embodiment of the present disclosure will be referred to as the PPG sensor 10 .
- FIG. 1 is a schematic configuration diagram of a pixelated monolithic photoplethysmography sensor having a spiking neural network structure according to an embodiment of the present disclosure.
- the PPG sensor 10 may include a plurality of input neurons 100 , a plurality of output neurons 200 and a plurality of synchronizers 300 .
- the PPG sensor 10 may include a plurality of input neurons 100 each composed of light sensing neurons (LSNs).
- LSNs light sensing neurons
- reference numeral 100 in the description of the embodiments of the present application may be used interchangeably for the light measuring neuron (LSN) 100 and the input neuron 100.
- input neurons 100 may be arranged in at least one of a column direction and a row direction.
- the plurality of input neurons 100 may be arranged to form an array including n x m pixels, but is not limited thereto.
- Column direction may be a horizontal direction (3 o'clock direction to 9 o'clock direction) based on the matters shown in FIG. ) can be.
- this direction setting may vary depending on the arrangement state of the present device.
- the PPG sensor 10 disclosed in the present disclosure may be arranged such that the row direction of FIG. 1 faces a horizontal direction or an oblique direction and the column direction faces a vertical direction or an oblique direction.
- each input neuron 100 may receive light and generate a spike signal proportional to the magnitude of the received light. A more detailed description of the input neuron 100 will be described later with reference to FIGS. 2 and 3 .
- the generated spike signal may be output to a column input node CI in a column direction or may be output to a row input node RI in a row direction.
- each of the plurality of input neurons 100 may be arranged to be connected to one column input node CI and one row input node RI.
- the column input node CI may include n nodes to which spike signals generated from m pixels corresponding to each column of the above-described array are applied
- the row input node (RI) may include m nodes to which spike signals generated from n pixels corresponding to each row of the above-described array are applied.
- the column input node CI includes CI[0], CI[1], ..., CI[n] of FIG. 1
- the row input node RI includes RI[0] of FIG. ], RI[1], ... , RI[m], but is not limited thereto.
- the PPG sensor 10 may include a plurality of output neurons 200 each of which is an output neuron (ON).
- output neuron ON
- reference numeral 200 in the description of the embodiments of the present application may be used interchangeably for ON 200 and output neuron 200 .
- the plurality of output neurons 200 include n output neurons 200 corresponding to any one of the columns constituting the array and m output neurons corresponding to any one of the rows constituting the array, respectively. (200) may be included.
- the total number of output neurons 200 may be 24, including 12 neurons arranged in a column direction and 12 neurons arranged in a row direction, but is not limited thereto.
- CO[0], CO[1], ... , CO[n] denote output terminals of the output neurons 200 receiving the spike signal from the column input node CI.
- RO[0], RO[1], ... , RO[n] may mean an output terminal of the output neuron 200 receiving the spike signal from the row input node RI.
- each output neuron 200 may sum spike signals applied from the plurality of input neurons 100 in a column direction or a row direction. That is, the output neuron 200 may sum spike signals input from a plurality of column input nodes CI or spike signals input from a plurality of row input nodes RI. In the case of the above example, any one output neuron 200 performing the summing based on the column direction may sum the spike signals applied from the 12 input neurons 100, but is not limited thereto.
- the output neurons 200 may generate an output spike signal when a result of summing the spike signals applied from the input neurons 100 along the row direction or the column direction is equal to or greater than a preset reference value. there is. That is, the output neuron 200 may output an output spike signal having an intensity proportional to the magnitude of light received by the input neuron 100 . A more detailed description of the output neuron 200 will be described later with reference to FIG. 4 .
- the PPG sensor 10 disclosed herein synchronizes the output spike signal with a digital signal It may include a synchronizer 300 for processing.
- each synchronizer (SYNC) 300 may be provided to be connected to an output terminal of each output neuron 200 .
- SYNC synchronizer
- COS[0], COS[1], ... , COS[n] represent output spike signals from CO[0], CO[1], ..., CO[n]. It may mean an output terminal of the synchronizer 300 that has been received, and ROS[0], ROS[1], ... , ROS[n] are RO[0], RO[1], ... , RO [n] may refer to an output stage of the synchronizer 300 receiving the output spike signal.
- the synchronizer 300 may receive output spike signals from the output neurons 200 and convert each received output spike signal into a digital signal. Specifically, the synchronizer 300 may receive a clock signal and convert an output spike signal into a 1-bit digital signal synchronized with the received clock signal. A more detailed description of the synchronizer 300 will be described later with reference to FIG. 5 .
- the synchronizer 300 may convert each output spike signal proportional to the magnitude of light received by the plurality of input neurons 100 into a digital signal.
- any input neuron 100 among a plurality of input neurons 100 is specified as a first LSN 100a
- the 1st LSN 100a and the 2nd LSN 100b can receive light of different sizes, respectively, and each has a different size proportional to the size of the received light.
- An intensity spike signal can be generated.
- the output neurons 200 receiving each spike signal may also be signals of different magnitudes (strengths).
- the PPG sensor 10 outputs a spike signal (or output spike signal) according to the spatial information of each pixelated light measurement neuron (LSN) 100 and a digital signal obtained by converting the spike signal.
- the PPG sensor 10 has a pattern of light applied with different intensities to each of the plurality of input neurons 100 (for example, the intensity of light incident on input neurons 100 relatively close to a light source is A plurality of outputs in which information on characteristics having a large value compared to the intensity of light incident on other input neurons 100 relatively far from the light source relative to the corresponding input neuron 100) is spatially summed in the column or row direction Since data is generated, it can be used to grasp specific spatial information of incident light applied to each input neuron (LSN) in units of pixels constituting the PPG sensor 10 through analysis of each output.
- the PPG sensor 10 may derive relative spatial information for the photoplethysmography sensor of the light applied to the plurality of input neurons 100 based on the magnitude information of each digital signal converted by the synchronizer 300.
- the PPG sensor 10 can easily remove noise caused by ambient light by summing and outputting a plurality of outputs according to spatial information of light in a matrix or vector form.
- FIG. 2 is a diagram showing a schematic circuit diagram of an input neuron according to an embodiment of the present disclosure.
- the input neuron 100 includes a Light-to-Digital Converter (LDC) 110, a first transistor 120, and a second transistor 130.
- LDC Light-to-Digital Converter
- the optical-to-digital converter 110 may receive light and generate a spike signal based on the magnitude of the received light. Specifically, the optical-to-digital converter 110 may generate a spike signal proportional to the magnitude of light.
- the first transistor 120 may output a spike signal (LDCo: LDC output) generated by the optical-to-digital converter 110 to a column input node CI corresponding to a column direction.
- the second transistor 130 may output the spike signal generated by the optical-to-digital converter 110 to a row input node RI corresponding to a row direction.
- the first transistor 120 and the second transistor 130 are illustrated as N Channel MOSFETs, but are not limited thereto.
- FIG. 3 is a diagram showing a circuit diagram of an optical-to-digital converter according to an embodiment of the present disclosure.
- the optical-to-digital converter 110 may include a photo diode 111 , a first comparator 112 , and a first reset MOSFET M RST , 113 .
- the photodiode 111 may convert light energy into electrical energy, and V PD may mean an output voltage by the photodiode 111 .
- the cathode terminal of the photodiode 111 may be grounded and the anode terminal may not be biased, but is not limited thereto.
- the first comparator 112 compares the reference voltage (V REF ) and the output voltage (V PD ) of the photodiode 111 to obtain an output voltage (V PD ) of the photodiode 111 . If the output voltage (V PD ) is less than the reference voltage (V REF ), a spike signal ( LDCo ) of a positive value or a spike signal (LDCo) is output. It can be output, but is not limited thereto.
- the first reset MOSFET 113 generates a value applied to the non-inverting terminal (+ terminal) of the first comparator 112 based on the spike signal LDCo applied to the gate GATE. can be reset.
- the first reset MOSFET 113 is shown as an N Channel MOSFET, but is not limited thereto.
- the first reset MOSFET 113 is an N Channel MOSFET
- the spike signal LDCo is a positive value
- a positive value is applied to the gate GATE of the first reset MOSFET 113
- the non-inverting terminal (+ terminal) of the first comparator 112 may conduct with the ground terminal, but is not limited thereto.
- FIG. 4 is a diagram illustrating a circuit diagram of an output neuron according to an exemplary embodiment of the present disclosure.
- the output neuron (ON) 200 may receive a summation result of spike signals output from a plurality of column input nodes (CI) or a plurality of row input nodes (RI). In other words, the output neuron 200 may sum the spike signals in a column direction or a row direction.
- the sum of the spike signals input to the output neurons 200 is specified as an input signal RI/CI.
- the output neuron 200 may include a second comparator 210 and a second reset MOSFET (M R , 220).
- the second comparator 210 receives a preset reference value ( VR_ON ) and an input signal (RI/CI), and receives the input reference value ( VR_ON ) and the input signal (RI/CI). ) can be compared to generate an output spike signal (RO/CO).
- the output spike signal RO/CO may be output as a positive value, and conversely, the input signal RI/CI may have the reference value If less than (V R_ON ), the output spike signal RO/CO may be output as a negative value or 0, but is not limited thereto.
- the output value of the second comparator 210 is inverted through the NOT gate, and the inverted signal may be applied to the gate GATE of the second reset MOSFET 220 . That is, the signal applied to the gate (GATE) of the second reset MOSFET 220 may be an inverted value (RO'/CO') of the output spike signal.
- the second reset MOSFET 220 outputs a non-inverting terminal ( + terminal) can be reset.
- the second reset MOSFET 220 is shown as a P Channel MOSFET, but is not limited thereto.
- the second reset MOSFET 220 is a P Channel MOSFET and the inverted output spike signal RO'/CO' has a negative value
- a negative value is applied to the gate of the second reset MOSFET 220.
- the first reset MOSFET 113 is turned on, and the non-inverting terminal (+ terminal) of the second comparator 210 may conduct with the ground terminal, but is not limited thereto.
- FIG. 5 is a diagram showing a circuit diagram of a synchronizer according to an embodiment of the present invention.
- synchronizer 300 may include a first flip-flop 310 , a second flip-flop 320 , a third flip-flop 330 and an XOR gate 340 .
- the first flip-flop 310, the second flip-flop 320, and the third flip-flop 330 may be, for example, D-flip-flops that operate by being triggered by a rising edge of an input clock. However, it is not limited thereto.
- an output spike signal RO/CO is input to a clock terminal, and an input terminal and an inverted output terminal may be connected. Accordingly, since the input terminal and the inverted output terminal of the first flip-flop 310 are connected to each other, the output is inverted according to the generation of the output spike signal RO/CO.
- the clock signal CLKS is input to the clock terminal of the second flip-flop 320 , and the input terminal may be connected to the output terminal of the first flip-flop 310 . Since the output of the second flip-flop 320 is determined according to the clock signal CLKS, it serves to synchronize the output spike signal RO/CO, whose occurrence time was random, to the clock signal CLKS, which is a reference clock signal.
- the clock signal CLKS is input to the clock terminal of the third flip-flop 330 , and the input terminal may be connected to the output terminal of the second flip-flop 320 .
- the third flip-flop 330 delays the output of the second flip-flop 320 by one clock to generate a signal for detecting the subsequent XOR gate 340 .
- the XOR gate 217 may receive inputs from the output terminals of the second flip-flop 320 and the output terminals of the third flip-flop 330 and output a 1-bit digital signal. Since the output signal of the second flip-flop 320 and the output signal of the third flip-flop 330 have a clock difference based on the clock signal CLKS, it is possible to generate a digital signal with a constant level.
- FIG. 6 is a graph exemplarily showing photoplethysmography sensor data and noise obtained using a pixelated monolithic photoplethysmography sensor according to an embodiment of the present disclosure.
- the photoplethysmography sensor (PPG sensor) data includes different output voltages for each magnitude of received light.
- the higher the output voltage eg, higher DC value or AC value
- LSN pixelated light measurement neuron
- FIG. 7 is a graph exemplarily illustrating a result of removing noise caused by ambient light using spatial information using photoplethysmography sensor data according to an embodiment of the present disclosure.
- the PPG sensor 10 may convert each output voltage that is individually output into an output from which noise due to ambient light is removed through a simple matrix operation without a conventional complex signal processing technique. Specifically, referring to the 'Ambient noise removed' section shown in FIG. 7, single output data (row direction and/or column direction) from which the effect of noise present in the 6 to 8 second section of the graph of FIG. 6 has been removed. It can be seen that data obtained by merging a plurality of output data derived respectively into one sequence) can be derived.
- the PPG sensor 10 may obtain an output from which noise due to ambient light is removed based on spatial information of light and a plurality of outputs.
- each output generated by the PPG sensor 10 eg, a voltage value corresponding to a digital signal output for each row or column of the pixel array) etc.
- a section in which the output value rapidly changes compared to the preceding and preceding sequences may be detected (detected) due to noise caused by ambient light interference, etc.
- the PPG sensor 10 includes such noise.
- a predetermined correction process to output data, a normal output from which noise is removed for a corresponding section may be derived.
- the PPG sensor 10 defines a plurality of outputs generated corresponding to incident light as an output matrix corresponding to a pixel array, and performs a matrix multiplication operation with a reference matrix for removing noise with respect to the corresponding output matrix. It may be to mount (have) a correction module (not shown) that operates to derive a normal output from which noise is removed in a method.
- the total output of the PPG sensor 10 is It can be merged in the form of a 12 X 1 matrix, and the correction module (not shown) of the PPG sensor 10 is a value capable of removing noise for the above-described abnormal section (the ambient noise section in FIG. 6), and each element is Normal output data of the corresponding section may be obtained by performing a matrix multiplication operation on the determined 1 ⁇ 12 size reference matrix with respect to the output matrix.
- the correction module multiplies and calculates a reference matrix in the form of a transposed matrix for an output matrix defined based on a plurality of outputs of the PPG sensor 10, thereby affecting the effect of unexpected noise reflected in an arbitrary section ( For example, interference due to ambient light, etc.) can be eliminated.
- the PPG sensor 10 disclosed herein estimates spatial information corresponding to the row direction and column direction A process of removing noise in some sections by utilizing spatial information reflected in the output data through matrix operation or vector operation on the plurality of output data since it provides multiple output data and can merge them in the form of a matrix or vector. There is an advantage that can be easily done.
- FIG. 8 is an operation flowchart of a method of driving a pixelated monolithic photoplethysmography sensor having a spiking neural network structure according to an embodiment of the present disclosure.
- the driving method of the pixelated monolithic photoplethysmography sensor of the spiking neural network structure shown in FIG. 8 may be performed by the PPG sensor 10 described above. Therefore, even if omitted below, the description of the PPG sensor 10 can be equally applied to the description of the driving method of the pixelated monolithic photoplethysmography sensor of the spiking neural network structure.
- the plurality of input neurons 100 may generate spike signals based on the magnitude of received light.
- the optical-to-digital converter 110 of each input neuron 100 may generate a spike signal having an intensity proportional to the magnitude of the received light.
- the spike signal generated by each input neuron 100 in step S11 may be output to a column input node corresponding to a column direction and a row input node corresponding to a row direction.
- each of the plurality of output neurons 200 may sum the spike signals applied from the plurality of input neurons 100 in a column direction or a row direction.
- each of the plurality of output neurons 200 may generate an output spike signal if the summation result in step S12 is equal to or greater than a predetermined reference value.
- the synchronizer 300 may convert each output spike signal output from each of the plurality of output neurons 200 into a digital signal.
- the synchronizer 300 may convert the output spike signal into a 1-bit digital signal synchronized with the clock signal.
- steps S11 to S14 may be further divided into additional steps or combined into fewer steps, depending on an embodiment of the present invention. Also, some steps may be omitted if necessary, and the order of steps may be changed.
- a driving method of a pixelated monolithic photoplethysmography sensor having a spiking neural network structure may be implemented in the form of program instructions that can be executed through various computer means and recorded in a computer readable medium.
- the computer readable medium may include program instructions, data files, data structures, etc. alone or in combination.
- Program instructions recorded on the medium may be specially designed and configured for the present invention, or may be known and usable to those skilled in computer software.
- Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic media such as floptical disks.
- - includes hardware devices specially configured to store and execute program instructions, such as magneto-optical media, and ROM, RAM, flash memory, and the like.
- program instructions include high-level language codes that can be executed by a computer using an interpreter, as well as machine language codes such as those produced by a compiler.
- the hardware devices described above may be configured to act as one or more software modules to perform the operations of the present invention, and vice versa.
- the above-described driving method of the pixelated monolithic photoplethysmography sensor of the spiking neural network structure may be implemented in the form of a computer program or application stored in a recording medium and executed by a computer.
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- General Health & Medical Sciences (AREA)
- Pathology (AREA)
- Heart & Thoracic Surgery (AREA)
- Veterinary Medicine (AREA)
- Public Health (AREA)
- Animal Behavior & Ethology (AREA)
- Surgery (AREA)
- Medical Informatics (AREA)
- Signal Processing (AREA)
- Cardiology (AREA)
- Physiology (AREA)
- Artificial Intelligence (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Psychiatry (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Optics & Photonics (AREA)
- Theoretical Computer Science (AREA)
- Power Engineering (AREA)
- Data Mining & Analysis (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Evolutionary Computation (AREA)
- Computational Linguistics (AREA)
- Transforming Light Signals Into Electric Signals (AREA)
Abstract
Description
Claims (10)
- 스파이킹 신경망 구조의 픽셀화된 모놀리식 광혈류측정 센서에 있어서,각각이 스파이크 신호를 생성하고, 로우 방향 및 칼럼 방향 중 적어도 하나의 방향으로 배열되는 복수의 입력 뉴론;상기 복수의 입력 뉴론으로부터 인가되는 상기 스파이크 신호를 상기 칼럼 방향 또는 상기 로우 방향을 기준으로 합산하고, 합산한 결과가 기 설정된 기준값 이상이면 출력 스파이크 신호를 생성하는 복수의 출력 뉴론; 및상기 출력 스파이크 신호 각각을 디지털 신호로 변환하는 동기화기,를 포함하는 것을 특징으로 하는,광혈류측정 센서.
- 제1항에 있어서,상기 복수의 입력 뉴론 각각은,수신한 광의 크기에 기반하여 상기 스파이크 신호를 생성하는 광-디지털 컨버터;상기 스파이크 신호를 상기 칼럼 방향에 대응하는 칼럼 입력 노드로 출력하는 제1트렌지스터; 및상기 스파이크 신호를 상기 로우 방향에 대응하는 로우 입력 노드로 출력하는 제2트렌지스터,를 포함하는 것인,광혈류측정 센서.
- 제2항에 있어서,상기 광-디지털 컨버터는,수신한 광의 크기에 비례하는 세기의 상기 스파이크 신호를 생성하는 것인,광혈류측정 센서.
- 제1항에 있어서,상기 복수의 입력 뉴론은,n x m개의 픽셀을 포함하는 어레이를 이루도록 배치되고,상기 칼럼 입력 노드는,상기 어레이의 각각의 칼럼에 대응하는 m개의 픽셀로부터 생성된 상기 스파이크 신호가 인가되는 n개의 노드를 포함하고,상기 로우 입력 노드는,상기 어레이의 각각의 로우에 대응하는 n개의 픽셀로부터 생성된 상기 스파이크 신호가 인가되는 m개의 노드를 포함하는 것인,광혈류측정 센서.
- 제4항에 있어서,상기 복수의 출력 뉴론은,각각이 상기 어레이를 이루는 칼럼 중 어느 하나에 대응하는 n개의 출력 뉴론 및 각각이 상기 어레이를 이루는 로우 중 어느 하나에 대응하는 m개의 출력 뉴론을 포함하는 것인,광혈류측정 센서.
- 제1항에 있어서,상기 동기화기는,상기 출력 스파이크 신호를 클록 신호에 동기화된 1비트 디지털 신호로 변환시키는 것인,광혈류측정 센서.
- 제1항에 있어서,상기 디지털 신호 각각의 크기 정보에 기초하여 상기 복수의 입력 뉴론으로 인가되는 광의 상기 광혈류측정 센서에 대한 상대적인 공간 정보가 도출되는 것을 특징으로 하는 것인,광혈류측정 센서.
- 스파이킹 신경망 구조의 픽셀화된 모놀리식 광혈류측정 센서의 구동 방법에 있어서,복수의 입력 뉴론이 수신한 광의 크기에 기반하여 스파이크 신호를 생성하는 단계;상기 복수의 입력 뉴론으로부터 인가되는 상기 스파이크 신호를 칼럼 방향 또는 로우 방향을 기준으로 합산하는 단계;합산한 결과가 기 설정된 기준값 이상이면, 복수의 출력 뉴론이 출력 스파이크 신호를 생성하는 단계; 및상기 출력 스파이크 신호 각각을 디지털 신호로 변환하는 단계,를 포함하는 것을 특징으로 하는,광혈류측정 센서의 구동 방법.
- 제8항에 있어서,상기 스파이크 신호를 생성하는 단계는,광-디지털 컨버터가 수신한 광의 크기에 비례하는 세기의 상기 스파이크 신호를 생성하고, 상기 스파이크 신호를 상기 칼럼 방향에 대응하는 칼럼 입력 노드 및 상기 로우 방향에 대응하는 로우 입력 노드로 출력하는 것인,광혈류측정 센서의 구동 방법.
- 제8항에 있어서,상기 디지털 신호로 변환하는 단계는,상기 출력 스파이크 신호를 클록 신호에 동기화된 1비트 디지털 신호로 변환시키는 것인,광혈류측정 센서의 구동 방법.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| US18/687,077 US12413868B2 (en) | 2021-12-03 | 2022-10-13 | Pixelated monolithic photoplethysmographic sensor having spiking neural network structure and driving method therefor |
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| KR1020210172377A KR102579571B1 (ko) | 2021-12-03 | 2021-12-03 | 스파이킹 신경망 구조의 픽셀화된 모놀리식 광혈류측정 센서 및 그의 구동 방법 |
| KR10-2021-0172377 | 2021-12-03 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| WO2023101201A1 true WO2023101201A1 (ko) | 2023-06-08 |
Family
ID=86612539
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| PCT/KR2022/015490 Ceased WO2023101201A1 (ko) | 2021-12-03 | 2022-10-13 | 스파이킹 신경망 구조의 픽셀화된 모놀리식 광혈류측정 센서 및 그의 구동 방법 |
Country Status (3)
| Country | Link |
|---|---|
| US (1) | US12413868B2 (ko) |
| KR (1) | KR102579571B1 (ko) |
| WO (1) | WO2023101201A1 (ko) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160283842A1 (en) * | 2014-03-06 | 2016-09-29 | Progress, Inc. | Neural network and method of neural network training |
| US10579925B2 (en) * | 2013-08-26 | 2020-03-03 | Aut Ventures Limited | Method and system for predicting outcomes based on spatio/spectro-temporal data |
| US20200202206A1 (en) * | 2018-12-07 | 2020-06-25 | Commissariat à l'énergie atomique et aux énergies alternatives | Artificial neuron for neuromorphic chip with resistive synapses |
| KR20200133246A (ko) * | 2018-03-14 | 2020-11-26 | 크로놀라이프 | 다수의 신호들을 프로세싱하기 위한 시스템 및 방법 |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR101987399B1 (ko) | 2017-07-11 | 2019-06-12 | 한국과학기술원 | 높은 신호대잡음비를 갖는 광용적맥파 센서를 위한 저각도 광학투과필터의 제조 방법 및 이를 구비한 광용적맥파 센서 |
-
2021
- 2021-12-03 KR KR1020210172377A patent/KR102579571B1/ko active Active
-
2022
- 2022-10-13 US US18/687,077 patent/US12413868B2/en active Active
- 2022-10-13 WO PCT/KR2022/015490 patent/WO2023101201A1/ko not_active Ceased
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10579925B2 (en) * | 2013-08-26 | 2020-03-03 | Aut Ventures Limited | Method and system for predicting outcomes based on spatio/spectro-temporal data |
| US20160283842A1 (en) * | 2014-03-06 | 2016-09-29 | Progress, Inc. | Neural network and method of neural network training |
| KR20200133246A (ko) * | 2018-03-14 | 2020-11-26 | 크로놀라이프 | 다수의 신호들을 프로세싱하기 위한 시스템 및 방법 |
| US20200202206A1 (en) * | 2018-12-07 | 2020-06-25 | Commissariat à l'énergie atomique et aux énergies alternatives | Artificial neuron for neuromorphic chip with resistive synapses |
Non-Patent Citations (2)
| Title |
|---|
| GEON-MYEONG LEE: "Behavior of the spiking neuron model and training of the spiking neural network", JOURNAL OF INFORMATION SCIENCE, vol. 38, no. 2, 1 February 2020 (2020-02-01), pages 8 - 19, XP093069437 * |
| MOSTAFA RAHIMI AZGHADI; COREY LAMMIE; JASON K. ESHRAGHIAN; MELIKA PAYVAND; ELISA DONATI; BERNABE LINARES-BARRANCO; GIACOMO INDIVER: "Hardware Implementation of Deep Network Accelerators Towards Healthcare and Biomedical Applications", ARXIV.ORG, CORNELL UNIVERSITY LIBRARY, 201 OLIN LIBRARY CORNELL UNIVERSITY ITHACA, NY 14853, 28 April 2021 (2021-04-28), 201 Olin Library Cornell University Ithaca, NY 14853 , XP081924537, DOI: 10.1109/TBCAS.2020.3036081 * |
Also Published As
| Publication number | Publication date |
|---|---|
| KR20230083921A (ko) | 2023-06-12 |
| US20240380991A1 (en) | 2024-11-14 |
| US12413868B2 (en) | 2025-09-09 |
| KR102579571B1 (ko) | 2023-09-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| KR100833177B1 (ko) | 자동적으로 오프 셋을 조정할 수 있는 신호변환회로 및 그방법 | |
| WO2015008902A1 (en) | Fingerprint recognition sensor capable of sensing fingerprint using optical and capacitive method | |
| WO2011099759A2 (en) | Image sensor and operating method | |
| WO2020045946A1 (ko) | 영상 처리 장치 및 영상 처리 방법 | |
| WO2018120315A1 (zh) | 一种触控显示面板的驱动方法、驱动装置以及触控显示器 | |
| WO2018155738A1 (ko) | 복수의 광원을 활용하는 양자 노이즈 기반 난수생성장치 | |
| US4584609A (en) | Parallel-input/serial output CCD register with clocking noise cancellation, as for use in solid-state imagers | |
| WO2023101201A1 (ko) | 스파이킹 신경망 구조의 픽셀화된 모놀리식 광혈류측정 센서 및 그의 구동 방법 | |
| WO2018111011A1 (ko) | 이동 객체 탐지 시스템 및 방법 | |
| WO2015160143A1 (ko) | 광 검출기 | |
| WO2013042918A1 (ko) | 복수 개의 램프 신호를 사용하는 다중 스텝 구조의 adc 및 이를 이용한 아날로그-디지털 변환 방법 | |
| EP1641250B1 (en) | Signal read circuit | |
| WO2023038417A1 (ko) | 줌 히스토그래밍 tdc를 이용한 플래시 라이다 센서 | |
| WO2014014265A1 (ko) | 영상 처리 장치 및 가로줄 노이즈 개선 방법 | |
| WO2018062650A2 (ko) | 스테레오 매칭 장치 및 방법 | |
| WO2018088649A1 (ko) | 글린트 검출 방법 | |
| WO2017099308A1 (ko) | 컬러 필터 어레이 및 이를 이용한 이미지 센서 | |
| WO2012020884A1 (ko) | 씨모스 이미지 센서 및 그것의 동작 방법 | |
| WO2024257957A1 (ko) | 시선추적센서 | |
| WO2026095547A1 (ko) | Cmos 영상 센서 기반 동공 영역 측정장치 | |
| JPS62198257A (ja) | カラ−リニアイメ−ジセンサ | |
| WO2025258753A1 (ko) | 엑스선 디텍터 | |
| WO2024210659A1 (ko) | 이미지 센싱 장치 및 이미지 처리 방법 | |
| WO2023239164A1 (ko) | 영상 획득 장치 및 방법 | |
| WO2025198137A1 (ko) | 다채널 공정 모니터링 시스템 및 방법 |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| 121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 22901561 Country of ref document: EP Kind code of ref document: A1 |
|
| WWE | Wipo information: entry into national phase |
Ref document number: 18687077 Country of ref document: US |
|
| NENP | Non-entry into the national phase |
Ref country code: DE |
|
| 122 | Ep: pct application non-entry in european phase |
Ref document number: 22901561 Country of ref document: EP Kind code of ref document: A1 |
|
| WWG | Wipo information: grant in national office |
Ref document number: 18687077 Country of ref document: US |