APPARATUS, SYSTEM AND METHOD FOR PROCESSING OPTICAL COMPONENTS - Patent application

Abstract

The method comprises receiving optical output data of an optical device, providing the optical output data to a trained neural network configured to convert the optical output data into an optical performance metric, and running the trained neural network to convert the provided optical output data into an optical performance metric for the optical device.

Description

[Technical field]

[0001] (CROSS REFERENCE TO RELATED APPLICATIONS) This application claims priority to U.S. Provisional Application No. 62 / 930,681, filed November 5, 2019, and U.S. Provisional Application No. 63 / 057,572, filed July 28, 2020, the entireties of which are incorporated by reference herein as if fully set forth below.

[0002] (Technical field) Apparatus, system and method for processing optical components and for assessing the quality of optical components - Patents.com [Background technology]

[0003] Optical components can be used to predictably and desirably control the state of light and to detect its changed state. For example, optical components are commonly used to generate and transmit signals (e.g., data signals) at high speeds, and also to receive and detect optical signals at high speeds, enabling high data capacity optical communication systems. Optical components interact with light through reflection, refraction, diffraction, absorption, etc.

[0004] Advances in optical components, such as their ability to generate and detect complex optical waveforms representing advanced modulation formats, have enabled higher capacity optical communications links at higher throughput. These components are key to the success of data center networks, WDM optical fiber networks, and similar applications. Thus, as the use of optical components continues to grow, the precision in the optical waveforms generated or detected continues to increase, and the need to ensure the quality of the components continues to increase.

[0005] For example, the related art uses a transmission and dispersion eye closure quaternary (TDECQ) penalty method to qualify optical transmitters (e.g., 5-level pulse amplitude modulation or "PAM-4" transmitters). Because PAM-4 optical systems may implement signal processing methods at the receiver, TDECQ is a statistical method to evaluate optical transmitter performance independent of the receiver. These transmitters, like many transmitters envisioned for fiber optic communication systems, must transmit waveforms of sufficient quality to allow a corresponding receiver with dynamic signal processing to form a communication link with a specified performance. The standard TDECQ method numerically performs signal processing in the form of a reference feedforward equalizer (FFE). The reference feedforward equalizer (FFE) is intended to mitigate impairments that occur in this link, including those associated with the transmitter, the optical channel, and the reference receiver. Using this test method, the transmitter must meet certain performance criteria. However, while TDECQ is a useful figure of merit for optical transmitters, it is a computationally intensive iterative process and therefore requires significant time, power consumption, and memory to calculate the TDECQ value. Thus, there is a need for improved apparatus, systems, and methods that provide improved processing of optical components to enhance manufacturing qualification efforts and ensure post-installation performance.

[0006] A second example in the related art includes optical quadrature amplitude modulation (QAM) transmitters and receivers. QAM transmitters are similar to PAM-4 transmitters in that they generate waveforms with multiple amplitudes, but also use a specific phase of the optical signal to encode data. These components therefore generate and receive optical signals that are amplitude and phase modulated to encode data. Both QAM transmitters and receivers are also similar to PAM-4 components, but are capable of signal processing to ensure that high quality waveforms are generated and received. This includes the ability to process waveforms to mitigate or eliminate impairments in the transmitter, optical channel, or receiver. These optical components must also independently meet qualification requirements during manufacture and use, with methods that are highly efficient and accurate dynamic signal processing implemented. Thus, there is a need for improved systems and methods to provide performance metrics for optical components intended to generate and receive QAM style signals.

[0007] Aspects of the present disclosure are directed to solving these and other problems. According to some aspects of the present disclosure, apparatus, systems, and methods may provide computer systems with improved capabilities for quickly and efficiently processing and / or evaluating optical components. Summary of the Invention

[0008] According to one embodiment, there is provided a method comprising receiving optical output data of an optical device; providing the optical output data to a trained neural network configured to convert the optical output data into an optical performance metric; and executing the trained neural network to convert the provided optical output data into an optical performance metric for the optical device.

[0009] According to one embodiment, there is provided a method of creating an optical evaluation system, the method comprising the steps of generating a plurality of optical waveforms from one or more representative devices or systems, evaluating the plurality of optical waveforms to determine a respective performance qualification of the one or more representative devices or systems, and operating a neural network to train the neural network with the plurality of optical waveforms and each of the performance qualifications.

[0010] According to one embodiment, an inspection apparatus is provided, the inspection apparatus comprising: a transmitter configured to output an initial signal to a first optical device, the first optical device being configured to convey the initial signal; an input configured to receive the conveyed signal from the first optical device; at least one processor; and a memory storing instructions that, when executed by the at least one processor, control the at least one processor to: compare the initial signal with the conveyed signal to generate optical output data; provide the optical output data to a trained neural network configured to convert the optical output data into an optical device performance metric; generate optical output data for the optical device based on the conveyed signal; provide the optical output data to a trained neural network configured to convert the optical output data into an optical performance metric; and execute the trained neural network to convert the provided optical output data into an optical device performance metric for the first optical device.

[0011] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate several embodiments of the presently disclosed subject matter and are intended to explain the principles of the presently disclosed subject matter, but are not intended to limit the scope of the presently disclosed subject matter in any way. [Brief description of the drawings]

[0012] [Figure 1A]FIG. 1 illustrates an example environment for implementing aspects of the present disclosure. [Figure 1B] FIG. 1 illustrates an example environment for implementing aspects of the present disclosure. [Figure 2A] FIG. 1 illustrates an example environment for implementing aspects of the present disclosure. [Figure 2B] FIG. 1 illustrates an example environment for implementing aspects of the present disclosure. [Diagram 3] FIG. 1 illustrates an example convolutional neural network. [Figure 4] 1 is a flowchart illustrating optical component evaluation according to an aspect of the present disclosure. [Diagram 5] FIG. 1 illustrates neural network training according to an aspect of the present disclosure. [Figure 6] FIG. 1 illustrates a synthetic waveform generation system according to an aspect of the present disclosure. [Figure 7] FIG. 1 is a block diagram illustrating an exemplary computer system according to an aspect of the present disclosure. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0013] Although specific embodiments of the present disclosure are described in detail, it will be understood that other embodiments are possible. Thus, the scope of the present disclosure is not intended to be limited to the details of construction and the arrangement of parts set forth in the following description or illustrated in the drawings. Other embodiments of the present disclosure may be implemented or carried out in various ways. Also, in describing the embodiments, certain terms are used for the sake of clarity. Each term is intended to assume its broadest meaning as understood by those skilled in the art and to include all technical equivalents that operate in a similar manner to accomplish a similar purpose.

[0014] It should also be noted that, as used in the specification and the appended claims, the singular forms "a," "an," and "the" include plural references unless the context clearly dictates otherwise. Reference to a composition containing "a" component is intended to include the other components in addition to the specified element.

[0015] Herein, ranges may be expressed as from "about" or "approximately" or "near" one particular value and / or to "about" or "approximately" another particular value. When such a range is expressed, other exemplary embodiments include from the one particular value and / or to the other particular value.

[0016] Herein, the use of the terms "having," "having," "including," or "including" is intended to have the same meaning as the terms "comprising" or "comprises," without limitation, and does not exclude the presence of other structures, materials, or acts. Similarly, the use of the terms "can" or "may" is intended to be open-ended and reflects that a structure, material, or act is not required, but the absence of such a term is not intended to reflect that a structure, material, or act is required. To the extent that structures, materials, or acts are considered to be required at the time, they are identified as such.

[0017] It will also be understood that the reference to one or more method steps does not exclude the presence of additional method steps or intervening method steps between the steps explicitly identified. Moreover, although the term "step" is used herein to imply different aspects of the method used, this term should not be understood to imply any particular order between the various steps disclosed herein unless and until an ordering of individual steps is expressly required.

[0018] The components described below as constituting various elements of the present disclosure are for purposes of illustration and not limitation. Many suitable components performing the same or similar functions as the components described herein are intended to be within the scope of the present disclosure. Such other components not described herein may include, for example, but are not limited to, similar components developed after the development of the subject matter of the present disclosure.

[0019] Various exemplary embodiments are described below to facilitate an understanding of the principles and features of the present disclosure, however the disclosure is not so limited and may be applicable in other contexts.

[0020] As mentioned above, related art optical component evaluation requires an intensive and iterative process. Aspects of the present disclosure are directed to solving these and other problems. According to some aspects of the present disclosure, apparatus, systems, and methods may provide computer systems with improved capabilities to quickly and effectively process and / or evaluate optical components. For example, aspects of the present disclosure utilize machine learning to convert eye diagram images or real-time waveforms into evaluation values. In experiments, the inventors have found that implementing aspects of the present disclosure can increase the speed by up to 1000 times and enhance the operation of the evaluation system.

[0021] Aspects of the present disclosure are described below with reference to the drawings. FIG. 1A illustrates an example environment 100a in which aspects of the present disclosure may be implemented. The environment 100a is a block diagram illustrating a training system that trains a neural network 155a to associate optical output data (e.g., eye diagrams or waveforms) with component performance metrics. The environment 100a includes a reference transmitter 110a, an optical link 114, and a calibrated optical receiver 116 in a training environment 120a. The reference transmitter 110a is connected to a signal data channel of a data source 130. The reference transmitter 110a outputs an optical waveform carried over the optical link 114 to the calibrated optical receiver 116. The calibrated optical receiver 116 receives the optical waveform and converts it to an electrical waveform.

[0022] The data receiver 140 receives the electrical waveforms from the calibrated optical receiver 116 and digitizes and captures the waveforms. The data receiver 140 may include an analog to digital converter (AtoD) and may be, for example, a real-time scope or an equivalent time sampling scope. The data receiver 140 provides sampled waveforms corresponding to the optical waveforms to the controller 150, which algorithmically determines parameter values ​​for each particular waveform received. The waveforms are paired with the determined parameters and provided to an untrained neural network 155a as training samples.

[0023] In some cases, the controller 150 (e.g., a computer) can pre-process the waveform and convert it into an eye diagram, constellation, or other shape that represents useful features of the waveform. The controller 150 can then pair the converted waveform with the determined parameters and provide the combined pair to the neural network 155a as training samples. Similarly, the data receiver 140 may pre-process the waveform and convert it into an eye diagram, constellation, or other shape that represents useful features of the waveform. In such cases, the converted waveform (e.g., eye diagram or constellation) can be provided to the controller 150 and then provided to the untrained neural network 155a along with the paired determined parameters as training samples.

[0024] The controller 150 can configure the data source 130 and / or the training environment 120a to generate waveforms that represent a range of performance for the transmitter being tested. For example, the controller 150 can cause the data source 130 to output waveforms that cause the reference transmitter 110a to emulate real parts with various stresses. Additionally, the controller 150 can control the optical link 114 and / or the calibrated optical receiver 116 to emulate various stresses on the communication link (e.g., reflections, temperature, dispersion, and bandwidth limitations).

[0025] The optical link 114 may be a short (e.g., insignificant) length of fiber connecting the reference transmitter 110a to the calibrated optical receiver 116, but this is merely one example. In some cases, the optical link 114 may be, for example, an optical cable representing an actual fiber link between the transmitter and receiver. In some cases, the optical link 114 may be comprised of multimode fiber, single mode fiber, and / or a concatenation of different types of fiber. The optical link 114 may include passive and active components including, for example, connectors, switches, amplifiers, and filters. In some cases, the optical link 114 may be partially or wholly artificial to more easily introduce various filters or other stresses into the communication link of the training environment 120a.

[0026] 1B illustrates an example environment 100b in which aspects of the present disclosure may be implemented. Environment 100b is a block diagram illustrating a test system for evaluating a test transmitter 110b based on its associated optical output data (e.g., eye diagram or waveform). Environment 100b includes a test transmitter 110b, an optical link 114, and a calibrated optical receiver 116 in a test environment 120b. Reference transmitter 110b is connected to a signal data channel of data source 130. Reference transmitter 110b outputs an optical waveform carried over optical link 114 to calibrated optical receiver 116. Calibrated optical receiver 116 receives the optical waveform and converts it to an electrical waveform.

[0027] The data receiver 140 receives the electrical waveform from the calibrated optical receiver 116 and digitizes and captures the electrical waveform. The data receiver 140 may include an analog to digital converter (AtoD) and may be, for example, a real-time scope or an equivalent time sampling scope. The data receiver 140 provides the sampled waveform to the controller 150, which provides the waveform to the trained neural network 155b. The trained neural network 155b processes the waveform and transforms it to generate parameters for the test transmitter 110b.

[0028] In some cases, the controller 150 (e.g., a computer) can pre-process the waveform and convert it into an eye diagram, constellation, or other shape that represents useful features of the waveform. The controller 150 can provide the modified waveform to a neural network 155b for testing, which can convert its input into parameters of the test transmitter 110b. Similarly, the data receiver 140 can pre-process the waveform and convert it into an eye diagram, constellation, or other shape that represents useful features of the waveform. In such cases, the modified waveform (e.g., eye diagram or constellation) can be provided to the controller 150 and then provided to the trained neural network 155b for processing.

[0029] The controller 150 can configure the data source 130 and / or the test environment 120b to generate waveforms that represent a range of performance for the transmitter being tested. For example, the controller 150 can cause the data source 130 to transmit stressed waveforms to the test transmitter 110b in response to various test conditions. Additionally, the controller 150 can control the optical link 114 and / or the calibrated optical receiver 116 to emulate various stresses (e.g., reflections, temperature, and bandwidth limitations) on the communication link.

[0030] The optical link 114 may be a short (e.g., a small) length of fiber connecting the reference transmitter 110b to the calibrated optical receiver 116, but this is just one example. In some cases, the optical link 114 may be a long optical cable representing, for example, an actual fiber link between a transmitter and a receiver. In some cases, the optical link 114 may be comprised of multimode fiber or single mode fiber. The optical link 114 may include passive and active components including, for example, connectors, switches, amplifiers, and filters. In some cases, the optical link 114 may be partially or wholly artificial to more easily introduce various stresses into the communication link of the test environment 120b.

[0031] FIG. 2A illustrates an example environment 200a in which aspects of the disclosure may be implemented. The environment 200a is a block diagram illustrating a training system that trains a neural network 155a to associate optical output data (e.g., eye diagrams or waveforms) with component performance metrics. The environment 200a includes a training environment 220a connected to a signal data channel of a data source 130. The training environment 220a may include a communication link including one or more optical components that emulate the potential behavior of a component that will be subsequently tested. Examples of optical components may include components or products that transmit, receive, or transmit data signals at very high speeds, such as high-speed data communication and telecommunication network equipment such as switches and routers, and fiber optic systems. Fiber optic systems include components, links and / or channels, cables, chip-to-chip communication links, and optical transmitters, receivers, and / or transceivers.

[0032] The data receiver 140 receives electrical waveforms from the training environment 220a and digitizes and captures the waveforms. The data receiver 140 may include an analog-to-digital converter (AtoD) and may be, for example, a real-time scope or an equivalent-time sampling scope. The data receiver 140 provides the sampled waveforms to the controller 150. The controller 150 determines parameter values ​​for each particular waveform received using one or more algorithms. The waveforms are paired with the determined parameters and provided as training samples to the untrained neural network 155a.

[0033] In some cases, the controller 150 (e.g., a computer) can pre-process the waveform and convert it into an eye diagram, constellation, or other shape that represents useful features of the waveform. The controller 150 can then pair the converted waveform with the determined parameters and provide the combined pair to the neural network 155a as training samples. Similarly, the data receiver 140 may pre-process the waveform and convert it into an eye diagram, constellation, or other shape that represents useful features of the waveform. In such cases, the converted waveform (e.g., eye diagram or constellation) can be provided to the controller 150 and then provided to the untrained neural network 155a along with the paired determined parameters as training samples.

[0034] Controller 150 can configure data source 130 and / or training environment 220a to generate waveforms that represent a range of performance for the transmitter being tested. For example, controller 150 can cause data source 130 to output waveforms that cause environment 220a to emulate real parts with various stresses. Controller 150 can also control other portions of the optical communications link to emulate various stresses on the communications link (e.g., reflections, temperature, and bandwidth limitations).

[0035] FIG. 2B illustrates an example environment 200b in which aspects of the present disclosure may be implemented. The environment 200b is a block diagram illustrating a test system for evaluating a generic test optical component 220b in a test environment based on its associated optical output data (e.g., eye diagram or waveform). The environment 200b includes a test environment 220b connected to a signal data channel of a data source 130. The test component 220b in the test environment conveys an electrical waveform to a data receiver 140. As will be appreciated by those skilled in the art in view of the present disclosure, the test component may be one or more portions of an optical communication link. For example, the test optical component may include components or products that transmit, receive, or transmit data signals at very high speeds, such as high speed data communication and telecommunication network equipment such as switches and routers, and fiber optic systems. Fiber optic systems include components, links and / or channels, cables, chip-to-chip communication links, and optical transmitters, receivers, and / or transceivers.

[0036] The data receiver 140 receives the electrical waveform from the calibrated optical receiver 116 and digitizes and captures the electrical waveform. The data receiver 140 may include an analog to digital converter (AtoD) and may be, for example, a real-time scope or an equivalent time sampling scope. The data receiver 140 provides the sampled waveform to the controller 150, which provides the waveform to the trained neural network 155b. The trained neural network 155b processes the waveform and transforms it to generate parameters of the part under inspection.

[0037] In some cases, the controller 150 (e.g., a computer) may pre-process the waveform and convert it into an eye diagram, constellation, or other shape that represents useful features of the waveform. The controller 150 may provide the modified waveform to the neural network 155b for testing, which may convert its input into parameters of the part under test. Similarly, the data receiver 140 may pre-process the waveform and convert it into an eye diagram, constellation, or other shape that represents useful features of the waveform. In such cases, the modified waveform (e.g., eye diagram or constellation) may be provided to the controller 150 and then provided to the trained neural network 155b for processing.

[0038] Controller 150 can configure data source 130 and / or test environment 220b to generate waveforms that represent a range of performance for the transmitter being tested. For example, controller 150 can cause data source 130 to transmit stressed waveforms to test environment 220b in response to various test conditions. Additionally, controller 150 can control portions of optical test environment 220b other than the components under test to emulate various stresses (e.g., reflections, temperature, and bandwidth limitations) on the communications link.

[0039] FIG. 3 illustrates an example convolutional neural network 300 according to an embodiment of the present disclosure. The neural network 300 may be the neural network 155 implemented by the controller 150. The neural network 300 receives as input a waveform or eye diagram 301 and converts the waveform or eye diagram into a performance qualification for an optical component (e.g., the optical component 110). The neural network 300 may be a convolutional neural network. In some cases, the neural network 300 may be a one-dimensional neural network (e.g., for processing waveforms) or a two-dimensional neural network (e.g., for processing eye diagrams). Those skilled in the art will recognize the convolutional layer 155 as a "convolutional neural network" or "neural network", although a particular neural node structure 350 (e.g., a fully connected layer) follows the convolutional layer.

[0040] The neural network 300 may include multiple layers configured to extract and / or enhance features, which may compress data and do so iteratively. For example, the neural network 300 may include two or more layers (e.g., feature extraction layers) 310, 320, 330. Each extraction layer 310, 320, 330 may include a convolutional layer 314, 324, 334, a regularized linear unit 316, 326, 336, and a pooling layer 318, 328, 338, respectively. The first extraction layer 310 is an input layer 310 and may receive as input the waveform or eye diagram 301 or some other output of a pre-processing operation. The second convolutional layer 320 may receive as its input the output of the first layer 310 and perform secondary processing on the waveform or eye diagram 301. A third convolutional layer 330 may receive the output from the second convolutional layer 320 and perform tertiary processing on the waveform or eye diagram 301. Those skilled in the art will recognize that "convolutional layer" refers to multiple functions and / or separable layers within the processing layers 310, 320, and 330.

[0041] The convolutional and pooling layers of the second and third extraction layers 320 and 330 may operate on smaller data sets than the convolutional and pooling layers of the previous extraction layer. For example, the input convolutional layer 314 may have dimensions greater than 40×30 (e.g., 47×37) and have 30 or more filters, the second convolutional layer 324 may have dimensions greater than 15×15 (e.g., 17×17) and have 10 or more filters, and the third convolutional layer 334 may have dimensions greater than 5×5 (e.g., 7×7) and have 8 filters. Similarly, the input pooling layer 318 may have dimensions of about 41×31 and a stride of about 15, the second pooling layer 328 may have dimensions of about 13×13 and a stride of about 6, and the third pooling layer 338 may have dimensions of about 7×7 and a stride of about 4. As will be appreciated by those skilled in the art, the examples shown are for a two-dimensional CNN, but one-dimensional CNNs may alternatively be implemented. As will be appreciated by those skilled in the art, the pooling layers may be max pooling layers.

[0042] The output from the final convolutional extraction layers 310, 320, 330 is transmitted to a dropout layer 340, which randomly drops nodes during training to reduce overfitting and improve generalization error in the trained neural network 300. Following the dropout layer 340, a neural node structure 350 outputs a performance qualification assessment for the optical component. The neural node structure 350 can be, by way of non-limiting example, a recurrent neural network that can take information from the previous layer and map it to a real-valued scalar as a numerical value of a performance metric, or a classification layer that takes information from the previous layer and maps it to one of a number of predefined entities that the user set when training the neural network.

[0043] FIG. 4 is a flow chart 400 illustrating optical component evaluation according to one embodiment. The method may be performed, for example, by an inspection system, such as inspection system 100b or 200b including one or more components as shown and described in FIGS. 1B and 2B, by way of example only. The inspection system receives (410) optical output data of a first optical device (e.g., inspection transmitter 110b). The optical device may be, for example, an optical transmitter. The inspection system captures (420) the optical output data, for example, using an oscilloscope.

[0044] In some cases, the output data is received (410) and captured (420) using a test and measurement device. The test and measurement device may output an initial signal to a first optical device. For example, the initial signal may be intended to be transmitted by or through the first optical device. The output of the first optical device may then be read out from the first optical device. In some cases, the initial signal is an initial analog signal and the read out signal is a read out analog signal. The test and measurement system may first convert an initial digital signal to an initial analog signal and provide the initial analog signal to the optical device. Similarly, the output of the optical device may be a read out analog signal and the test and measurement machine may convert the read out analog signal to a read out digital signal prior to or as part of generating the optical waveform and / or eye diagram. In some cases, the input signal may be fixed for testing, but this is merely an example. In some cases, the input signal is variable and the test and measurement system may, for example, compare the initial signal to the read out signal to determine the optical output data. A variable input signal can be used to test different aspects or features of the optical device, or to test the optical device for different conditions. By comparing the input and output signals, the system can be adapted to change tests or conditions.

[0045] The captured optical output data can be an optical waveform. The inspection system provides the captured optical output data to the trained neural network (430). For example, the inspection system may provide the optical waveform directly to the neural network (430). However, this is merely an example. In some cases, the inspection system first pre-processes the waveform, for example, converting it into an eye diagram. The inspection system may then provide the eye diagram to the neural network (430).

[0046] A neural network is executed to convert the provided optical output data into an optical device performance qualification assessment for the first optical device (440). The neural network may be a convolutional neural network, where the provided optical output data corresponds to an optical waveform and the neural network may be a one-dimensional neural network, or where the provided optical output data corresponds to an eye diagram and the neural network may be a two-dimensional neural network. However, these are merely examples.

[0047] As is known, the performance of an optical component can be affected by the electrical components to which it is connected or contained. For example, optical transceivers often include electrical amplifiers and optical modulators. Thus, as will be appreciated by those of skill in the art in light of this disclosure, the electrical components can be tested separately from the optical portions of the optical component.

[0048] 5 is a flow chart 500 illustrating neural network training according to one embodiment. The method includes generating 510 an optical waveform from an optical device (e.g., a representative optical device). Generating the optical waveform includes placing the optical device in an inspection machine and causing the optical device to output a waveform corresponding to an operating range that represents a fault or non-ideal performance, and capturing the output from the device with an oscilloscope.

[0049] In some cases, generating an optical waveform may include generating a composite waveform corresponding to one or more virtual devices (e.g., virtual optical devices) or systems (e.g., optical communication systems). An optical communication system or optical communication link may include an optical transmitter, an optical receiver, an optical path (e.g., fiber optic path) connected therebetween, and related elements. This may include modeling a virtual optical device and generating a waveform based on the modeled device. The generated waveform corresponds to defined device parameters that control the virtual device performance. However, this is merely an example. In some cases, the composite waveform may include controlling a laser to output a signal, processing (e.g., modulating) this signal with a modulator, and capturing the modulated signal as a composite waveform with an oscilloscope. Multiple optical waveforms may be generated by varying the output of a data generator connected to the modulator. In some cases, a filter may be implemented across the modulated signal. The filter may emulate a bandwidth limit. In light of this disclosure, one of ordinary skill in the art will recognize that there are other methods for generating waveforms that represent desired performance ranges for optical devices and systems.

[0050] The method further includes evaluating (520) the optical waveforms to determine a respective performance qualification of the optical device. This can be done, for example, using the related art TDECQ method. If more waveforms are needed (425: Yes), new waveforms are then generated. If generating (510) the optical waveforms is done synthetically, the power provided to the modulators and / or the bandwidth of the filters may be adjusted to generate different waveforms.

[0051] Once enough waveforms and assessments have been accumulated (425: No), the neural network is operated to train the neural network with the multiple optical waveforms and respective performance qualification assessments (530). In some cases, before training the neural network, the waveforms are converted to eye diagrams. In this case, the neural network can be trained to process the waveforms directly or to process the eye diagrams. In some cases, the waveforms can be processed more quickly, but with slightly less fidelity.

[0052] In one embodiment, a neural network that has already been trained on other test attempts can be retrained using transfer learning techniques to account for changes in the new test attempt, such as changes in the particular analog-to-digital converters, cables, or other test hardware. Such transfer learning can reduce the time to retrain a new neural network while maintaining high performance.

[0053] FIG. 6 is a block diagram illustrating a synthetic waveform generation system 600 according to an embodiment. The synthetic waveform generation system 600 may generate a synthetic waveform for training a neural network (e.g., neural network 155 and / or 300). The system 600 may utilize a laser 610 that is modulated by a modulator 630 based on the output of a waveform generator 620. The waveform generator 620 may be an arbitrary waveform generator. Additionally, the system 600 may implement a Gaussian filter to emulate bandwidth limitations. Thus, the system 600 may capture the dispersive effects of the test fiber as well as the effects of a short length of single mode fiber. The modulator 630 may drive a voltage and / or vary the filter bandwidth to obtain multiple signal eye diagrams. The eye diagrams may be converted using an analog-to-digital converter 650. These signal eye diagrams may be evaluated (e.g., using the related art TDECQ method) to calculate an evaluation value. By doing this iteratively, a synthetic training set of eye diagrams and estimates can be generated for training and testing the neural network 660. By utilizing the synthetic waveform generation system 600, a wider range of estimates and eye diagram types can be utilized than would normally be available for a single training set, and thus in some cases the performance of the estimate system can be further improved.

[0054] 7 is a block diagram illustrating a computer system architecture 700 according to one example implementation. In light of the present disclosure, one of ordinary skill in the art will appreciate that aspects of the present disclosure can be implemented using one or more elements from computer system architecture 700. It will be appreciated that computer device architecture 700, for example, is provided for illustrative purposes only and is not intended to limit the scope of various implementations of the systems, methods, and computer-readable media disclosed herein.

[0055] The computing device architecture 700 of FIG. 7 includes a central processing unit (CPU) 702 for processing computer instructions and a display interface 704 for acting as a communication interface and providing functionality for displaying video, graphics, images, and text on a screen. In certain example implementations of the disclosed technology, the display interface 704 may be directly connected to a local display, such as a touch screen display associated with a mobile computing device. In another example implementation, the display interface 704 may be configured to provide data, images, and other information for an external / remote display 750 that is not necessarily physically connected to the mobile computing device. For example, a desktop monitor may be used to render graphics and other information presented on the mobile computing device. In certain example implementations, the display interface 704 may wirelessly communicate with the external / remote display 750, for example, via a Wi-Fi channel or other available network connection interface 712.

[0056] In one example implementation, the network connection interface 712 may be configured as a communications interface and may provide the ability to display video, graphics, images, text, other information, or any combination thereof on a screen. In one example, the communications interface may include a serial port, a parallel port, a general purpose input / output (GPIO) port, a game port, a universal serial bus (USB), a micro USB port, a high definition multimedia (HDMI) port, a video port, an audio port, a Bluetooth port, a near field communication (NFC) port, other similar communications interfaces, or any combination thereof. In one example, the display interface 704 may be operatively coupled to a local display, such as a touch screen display, associated with the mobile device. In another example, the display interface 704 may be configured to provide video, graphics, images, text, other information, or any combination thereof for an external / remote display 750 that is not necessarily connected to the mobile computing device. In one example, a desktop monitor may be used to depict or enhance graphical information that may be presented on the mobile computing device. In another example, the display interface 704 may communicate wirelessly with an external / remote display 750 via a network connection interface 712, such as a Wi-Fi transceiver.

[0057] The computing device architecture 700 may include a keyboard interface 706 that provides a communication interface to a keyboard. In one example implementation, the computing device architecture 700 may include a presence-sensitive display interface 708 for connecting to a presence-sensitive display 705. According to certain example implementations of the disclosed technology, the presence-sensitive display interface 708 may provide a communication interface to various devices, such as a pointing device, a touch screen, a depth camera, etc. These devices may or may not be associated with a display.

[0058] The computing device architecture 700 may be configured to allow a user to incorporate information into the computing device architecture 700 using input devices via one or more input / output interfaces (e.g., keyboard interface 706, display interface 704, presence-sensitive display interface 708, network connection interface 712, camera interface 714, sound interface 716, etc.). The input devices may include a mouse, a trackball, a directional pad, a trackpad, a touch-enabled trackpad, a presence-sensitive trackpad, a presence-sensitive display, a scroll wheel, a digital camera, a digital video camera, a webcam, a microphone, a sensor, a smart card, etc. Furthermore, the input devices may be integrated into the computing device architecture 700 or may be separate devices. For example, the input devices may be an accelerometer, a magnetometer, a digital camera, a microphone, and a light sensor.

[0059] An example implementation of the computing device architecture 700 may include an antenna interface 710 that provides a communication interface to an antenna, and a network connection interface 712 that provides a communication interface to a network. As described above, the display interface 704 may communicate with the network connection interface 712 to provide information for display, for example, on a remote display not directly connected or attached to the system. In certain example implementations, a camera interface 714 is provided that functions as a communication interface and provides the ability to capture digital images from a camera. In certain example implementations, a sound interface 716 is provided as a communication interface for converting sound to electrical signals using a microphone and electrical signals to sound using a speaker. In accordance with an example implementation, a random access memory (RAM) 718 is provided, in which computer instructions and data are stored in volatile memory and processed by the CPU 702.

[0060] According to one example implementation, the computing device architecture 700 includes a read-only memory (ROM) 720, in which unchanging low-level system code or data for basic system functions such as input and output (I / O), start-up, receiving keystrokes from a keyboard, etc., are stored in a non-volatile memory device. According to one example implementation, the computing device architecture 700 includes a storage medium 722 or other suitable type of memory (e.g., RAM, ROM, programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), magnetic disk, optical disk, floppy disk, hard disk, removable cartridge, flash drive). In this memory, files are stored, including an operating system 724, application programs 726 (including, as appropriate, e.g., a web browser application, a widget or gadget engine, and / or other applications), and data files 728. According to one example implementation, computing device architecture 700 includes a power supply 730 that provides a suitable alternating current (AC) or direct current (DC) to the power components.

[0061] According to one example implementation, the computing device architecture 700 includes a telephony communications subsystem 732 that enables the device 700 to send and receive sound over a telephone network. The components and the CPU 702 communicate with each other via a bus 734.

[0062] According to one exemplary implementation, the CPU 702 has a suitable structure to be a computer processor. In one configuration, the CPU 702 may include two or more processing units. The RAM 718 interfaces with the computer bus 734 and provides rapid RAM storage to the CPU 702 during execution of software programs, such as an operating system, application programs, and device drivers. More specifically, the CPU 702 loads computer-executable process steps from the storage medium 722 or other media into the fields of the RAM 718 and executes the software programs. Data may be stored in the RAM 718, where it may be accessed by the computer CPU 702 during execution.

[0063] The storage medium 722 itself may include a number of physical drive units. Examples of physical drive units are a redundant array of independent disks (RAID), a floppy disk drive, an external hard disk drive, a thumb drive, a pen drive, a key drive, a high-definition digital versatile disk (HD-DVD) optical disk drive, an internal hard disk drive, a Blu-Ray optical disk drive, or a holographic digital data storage (HDDS) optical disk drive, an external mini dual in-line memory module (DIMM) synchronous dynamic random access memory (SDRAM), or an external micro-DIMM SDRAM. Such computer-readable storage media enable a computer device to access computer-executable process steps, application programs, etc. stored on removable and non-removable memory media, and to unload data from or load data to the device. A computer program product, for example a computer program product utilizing a communication system, may be tangibly embodied in the storage medium 722. The storage medium 722 may include a machine-readable storage medium.

[0064] According to one example implementation, as used herein, the term computing device may be a CPU or may be conceptualized as a CPU (e.g., CPU 702 in FIG. 7). In this example implementation, the computing device (CPU) may be coupled, connected, and / or in communication with one or more peripheral devices, such as a display. In another example implementation, as used herein, the term computing device may refer to a mobile computing device, such as a smartphone, tablet computer, or smartwatch. In this example implementation, the computing device may output content to its local display and / or speaker. In another example implementation, the computing device may output content to an external display device, such as a TV or an external computer system (e.g., via Wi-Fi).

[0065] In example implementations of the disclosed technology, a computing device may include any number of hardware and / or software applications executing to facilitate certain operations. In example implementations, one or more I / O interfaces may facilitate communication between the computing device and one or more input / output devices. For example, a universal serial bus port, a serial port, a disk drive, a CD-ROM device, and / or one or more user interface devices (e.g., a keyboard, a keypad, a mouse, a control panel, a touch screen display, a microphone, etc.) may facilitate user interaction with the computing device. The one or more I / O interfaces may be used to receive or collect data and / or user commands from various input devices. The received data may be processed by one or more computer processors as desired in various implementations of the disclosed technology and / or may be stored in one or more memory devices.

[0066] One or more network interfaces may facilitate connection of the computing device inputs and outputs to one or more suitable networks and / or connections. An example of such a connection is a connection that facilitates communication with any number of sensors associated with the system. The one or more network interfaces may further facilitate connection to one or more suitable networks. An example of the one or more suitable networks is a local area network, a wide area network, the Internet, a cellular network, a radio frequency network, a Bluetooth enabled network, a Wi-Fi enabled network, any satellite based wired network, any wireless network, etc., for communicating with external devices and / or systems.

[0067] An embodiment of the present disclosure can be implemented based on at least the following.

[0068] Item 1: A method comprising: receiving optical output data of an optical device; providing the optical output data to a trained neural network configured to convert the optical output data into an optical performance metric; and executing the trained neural network to convert the provided optical output data into an optical performance metric for the optical device.

[0069] Item 2: The method of item 1, wherein the optical device comprises an optical transmitter.

[0070] Item 3: The method of item 1 or 2, wherein the performance metric relates to an optical communication system attached to the optical device.

[0071] Item 4: The method according to any one of items 1 to 3, wherein the performance metric relates to a Transmission and Dispersion Eye Closure Quaternary (TDECQ) penalty method.

[0072] Item 5: The method according to any one of items 1 to 4, wherein the optical device is an optical receiver.

[0073] Item 6: The method according to any one of items 1 to 5, wherein the optical output data includes a waveform of an output of the optical device.

[0074] Item 7: The method of any one of items 1 to 6, wherein the optical output data includes an eye diagram generated from a waveform of the output of the optical device, and the eye diagram is provided to the trained neural network.

[0075] Item 8: The method of item 7, further comprising the steps of receiving a waveform of the output of the optical device, and performing pre-processing on the waveform to convert the waveform into an eye diagram.

[0076] Item 9: The method of any one of items 1 to 8, wherein the trained neural network includes a convolutional neural network.

[0077] Item 10: The method of item 9, wherein the convolutional neural network includes multiple extraction layers configured to perform convolutional filtering, and the optical output data is fed to a first extraction layer and the input of a subsequent extraction layer is the output of a previous extraction layer.

[0078] Item 11: The method of item 9 or 10, wherein the convolutional neural network includes an input extraction layer having a convolutional filter and a pooling layer and to which the optical output data is supplied, a second extraction layer having a convolutional filter and a pooling layer and to which an output of the input extraction layer is supplied, and a third extraction layer having a convolutional filter and a pooling layer and to which an output of the second extraction layer is supplied.

[0079] Term 12: The method of term 11, wherein the pooling layer is a max pooling layer.

[0080] Item 13: The method according to any one of items 1 to 12, wherein the trained neural network comprises a two-dimensional convolutional neural network.

[0081] Item 14: The method of any one of items 1 to 12, wherein the trained neural network includes a one-dimensional convolutional neural network.

[0082] Item 15: The method according to any one of items 1 to 4, further comprising the step of capturing the light output data using an inspection and measurement device.

[0083] Item 16: The method according to item 15, wherein the inspection and measurement device includes a real-time scope.

[0084] Item 17: The method of item 15 or 16, wherein the inspection and measurement device includes an equivalent time scope.

[0085] Item 18: A method according to any one of items 1 to 17, wherein the step of capturing the optical output data includes a step of outputting an initial signal into the optical device, the initial signal being transmitted through the optical device, and a step of reading out the transmitted signal from the optical device.

[0086] Item 19: The method of item 18, wherein the initial signal is an initial analog signal and the read-out signal is a read-out analog signal, further comprising the steps of converting an initial digital signal to generate the initial analog signal, and converting the read-out analog signal to a read-out digital signal.

[0087] Item 20: The method of item 18 or 19, wherein the initial signal is fixed.

[0088] Item 21: The method of item 18 or 19, wherein the initial signal is variable.

[0089] Item 22: The method according to any one of items 1 to 21, wherein the step of acquiring the light output data further comprises a step of comparing the initial signal and the read-out signal.

[0090] Item 23: A method of creating an optical evaluation system, comprising the steps of generating a plurality of optical waveforms from one or more representative devices or systems, evaluating the plurality of optical waveforms to determine a respective performance qualification of the one or more representative devices or systems, and operating a neural network to train the neural network with the plurality of optical waveforms and each of the performance qualifications.

[0091] Clause 24: The method described in Clause 23, wherein the step of generating the multiple optical waveforms includes, for each device of the one or more representative devices or systems, a step of placing the representative device in an inspection machine and a step of capturing output from the representative device or system using an oscilloscope.

[0092] Clause 25: The method of clause 23 or 24, wherein the step of generating a plurality of optical waveforms includes the step of generating a composite waveform corresponding to one or more virtual representative devices or systems.

[0093] Clause 26: The method of clause 25, wherein the step of generating a composite waveform comprises the steps of modeling the one or more virtual representative devices and generating a plurality of waveforms based on the modeled devices. The plurality of waveforms can represent a range of performance.

[0094] Item 27: The method of item 25 or 26, further comprising the steps of controlling a laser to output a signal, processing the signal using a modulator, and capturing the modulated signal as a composite waveform using an oscilloscope.

[0095] Clause 28: The method of clause 27, wherein the modulator is adjusted based on the output of a waveform generator.

[0096] Clause 29: The method of clause 28, further comprising adjusting a power supply unit to the modulator to generate a plurality of different waveforms.

[0097] Clause 30: The method of any one of clauses 27 to 29, further comprising implementing a filter across the modulated signal.

[0098] Clause 31: The method of clause 30, further comprising adjusting a bandwidth of the filter to generate a plurality of different waveforms.

[0099] Item 32: An inspection apparatus comprising a transmitter configured to output an initial signal to an optical device, the optical device comprising a transmitter configured to convey the initial signal, an input configured to receive the signal conveyed from the optical device, at least one processor, and a memory storing instructions, the instructions, when executed by the at least one processor, control the at least one processor to generate optical output data of the optical device based on the conveyed signal, provide the optical output data to a trained neural network configured to convert the optical output data into an optical performance metric, and execute the trained neural network to convert the provided optical output data into an optical performance metric for the optical device.

[0100] Item 33: The system of item 32, wherein the optical device includes an optical transmitter.

[0101] Item 34: The system of items 32 or 33, wherein the performance metric relates to an optical communication system attached to the optical device.

[0102] Clause 35: The system of any one of clauses 32 to 34, wherein the performance metric relates to a Transmission and Dispersion Eye Closure Quaternary (TDECQ) penalty method.

[0103] Item 36: The system described in any one of items 32 to 35, wherein the optical device is an optical receiver.

[0104] Item 37: The system described in any one of items 32 to 36, wherein the optical output data includes a waveform of the output of the optical device.

[0105] Item 38: The system of any one of items 32 to 37, wherein the optical output data includes an eye diagram generated from a waveform of the output of the optical device, and the eye diagram is provided to the trained neural network.

[0106] Clause 39: The system described in any one of clauses 32 to 38, wherein the instructions further control the at least one processor to generate a waveform of the output of the optical device, perform pre-processing on the waveform, and convert the waveform into an eye diagram.

[0107] Item 40: The system of any one of items 32 to 39, wherein the trained neural network includes a convolutional neural network.

[0108] Item 41: The system described in Item 40, wherein the convolutional neural network includes a plurality of extraction layers configured to perform convolutional filtering, the optical output data being fed to the extraction layers and the input of a subsequent extraction layer being the output of a previous extraction layer.

[0109] Item 42: The system described in Item 40 or 41, wherein the convolutional neural network includes an input extraction layer having a convolutional filter and a pooling layer and to which the optical output data is supplied, a second extraction layer having a convolutional filter and a pooling layer and to which the output of the input extraction layer is supplied, and a third extraction layer having a convolutional filter and a pooling layer and to which the output of the second extraction layer is supplied.

[0110] Item 43: The system of item 42, wherein the pooling layer is a max pooling layer.

[0111] Item 44: The system of any one of items 32 to 43, wherein the trained neural network includes a two-dimensional convolutional neural network.

[0112] Item 45: The system of any one of items 32 to 44, wherein the trained neural network includes a one-dimensional convolutional neural network.

[0113] Item 46: The system according to any one of items 32 to 45, wherein the input unit includes a real-time scope.

[0114] Item 47: The system according to any one of items 32 to 46, wherein the input unit includes an equivalent time scope.

[0115] Item 48: The system according to any one of items 32 to 47, wherein the initial signal is fixed.

[0116] Item 49: The system described in any one of items 32 to 47, wherein the initial signal is variable.

[0117] Item 50: The system of any one of items 32 to 49, wherein capturing the optical output data further includes comparing the initial signal to the carried signal.

[0118] A system comprising at least one processor and one or more memories storing instructions which, when executed by the at least one processor, control the at least one processor to perform a method according to any one of claims 1 to 31.

[0119] A non-transitory computer-readable medium having stored thereon instructions for carrying out the method according to any one of claims 1 to 31.

[0120] In addition to the specification and drawings provided, the present disclosure includes an appendix that describes details of the development of the device according to the present disclosure, which is for illustrative purposes only and not for limiting purposes.

[0121] Although the present disclosure has been described in connection with several exemplary embodiments, as shown in various figures and described in the specification above, it will be understood that other similar embodiments can be used and modifications and additions can be made to the described embodiments to perform the same functions of the present disclosure without departing from the same. For example, in various embodiments of the present disclosure, methods and compositions have been described in accordance with the subject matter disclosed herein. However, other methods or compositions equivalent to the described embodiments are also contemplated according to the present teachings. Thus, the present disclosure should not be limited to any single embodiment, but rather should be construed in breadth and scope in accordance with the appended claims.

Claims

1. providing light output data for the optical device to a trained neural network configured to convert the light output data into an optical performance metric; and executing the trained neural network to convert the provided light output data into optical performance metrics for the optical device. The performance metric is for a transmit and variance eye closure quaternary (TDECQ) penalty method; The optical output data is a waveform of the output of the optical device; an eye diagram generated from a waveform at the output of the optical device; or a constellation generated from a waveform of the output of the optical device.

2. receiving the optical output data of the optical device; The method of claim 1 , wherein the optical device comprises a device selected from the group consisting of an optical transmitter and an optical receiver.

3. The method of claim 1 or claim 2, wherein the performance metric further relates to an optical communication system attached to the optical device.

4. Further comprising the step of feeding the eye diagram to the trained neural network, The method of claim 1 , wherein the optical output data includes the eye diagram generated from a waveform of the output of the optical device.

5. receiving a waveform from an output of the optical device; The method of claim 1 , further comprising: performing pre-processing on the waveform to convert the waveform into the eye diagram.

6. providing light output data relating to the optical device to a trained neural network; converting, by the trained neural network, the light output data into one or more optical performance metrics for the optical device; the trained neural network is selected from the group consisting of a convolutional neural network, a two-dimensional convolutional neural network, and a one-dimensional convolutional neural network; at least one of the optical performance metrics is a transmit and distributed eye closure quaternary (TDECQ) metric; The optical output data is a waveform of the output of the optical device; an eye diagram generated from a waveform at the output of the optical device; or a constellation generated from a waveform of the output of the optical device.

7. the trained neural network includes the convolutional neural network; the convolutional neural network includes a plurality of abstraction layers configured to perform convolutional filtering; the optical output data is first provided to a first extraction layer; The method of claim 6 , wherein the inputs of each subsequent extraction layer include the outputs of each previous extraction layer.

8. the trained neural network includes the convolutional neural network; The convolutional neural network includes: an input extraction layer comprising a convolution filter and a pooling layer, the input extraction layer being fed with the optical output data; a second extraction layer having a convolution filter and a pooling layer and receiving the output of the input extraction layer; a third extraction layer having a convolution filter and a pooling layer, the third extraction layer being fed with an output of the second extraction layer.

9. The method of claim 8 , wherein the pooling layer is a max pooling layer.

10. The method of any of claims 6 to 9, further comprising capturing the light output data with a test and measurement device.

11. 11. The method of claim 10, wherein the test and measurement device comprises a device selected from the group consisting of a real-time scope and an equivalent-time scope.

12. The step of acquiring the light output data includes: outputting an initial analog signal from the test and measurement device into the optical system; outputting an analog signal transmitted from the optical device; and reading out the analog signal transmitted from the optical device.

13. converting an initial digital signal to generate said initial analog signal; 13. The method of claim 12, further comprising the step of converting the transmitted analog signal into a read-out digital signal.

14. The method of claim 12 or 13, wherein the initial analog signal is selected from the group consisting of a fixed initial analog signal and a variable initial analog signal.

15. 15. The method of claim 14, wherein the step of capturing light output data further comprises the step of comparing the initial analog signal to the transmitted analog signal.

16. The method of claim 1, further comprising generating the light output data from one or more representative devices or systems; the providing step includes providing the light output data from the one or more representative devices or systems to the trained neural network; The optical output data is a plurality of optical waveforms from said one or more representative devices or systems; an eye diagram generated from the optical waveforms from the one or more representative devices or systems; or a constellation generated from the plurality of optical waveforms from the one or more representative devices or systems; 2. The method of claim 1, wherein the converting step includes evaluating one or more of the plurality of optical waveforms with the trained neural network to determine the optical performance metric for at least one of the representative devices or systems.

17. The step of generating light output data includes, for each representative device or system of the one or more representative devices or systems: placing the representative device or system in an inspection machine; 17. The method of claim 16, comprising the step of capturing an output from the representative device or system with an oscilloscope.

18. 20. The method of claim 16, wherein generating the light output data comprises generating a synthetic waveform corresponding to one or more hypothetical representative devices or systems.

19. The step of generating a composite waveform includes: modeling one or more of the virtual representative devices or systems; and generating a waveform based on the modeled device or system.

20. controlling a laser to output a signal; processing said signal with a modulator; 20. The method of claim 18 or 19, further comprising the step of capturing the modulated signal as one of the composite waveforms with an oscilloscope.

21. 21. The method of claim 20, further comprising adjusting the modulator based on an output of a waveform generator.

22. 22. The method of claim 20 or claim 21, further comprising adjusting a power supply unit to the modulator to generate a plurality of different waveforms.

23. 23. The method of any of claims 20 to 22, further comprising the step of cross filtering the modulated signal.

24. 24. The method of claim 23, further comprising adjusting a bandwidth of the filtering to generate a plurality of different waveforms.

25. 13. An inspection system for carrying out the method of claim 1, comprising: a transmitter configured to output an initial signal to the optical device configured to carry the initial signal; an input configured to receive the signal conveyed from the optical device; At least one processor; A memory storing an instruction; The instructions, when executed by the at least one processor, generating the optical output data of the optical device based on the conveyed signal; providing the light output data to the trained neural network configured to convert the light output data to the optical performance metric; executing the trained neural network to convert the provided light output data into the optical performance metric qualification assessment for the optical device; An inspection device that controls the at least one processor.

26. 26. The system of claim 25, wherein the optical output data includes an eye diagram generated from a waveform of an output of the optical device, the eye diagram being provided to the trained neural network.

27. The instruction: generating a waveform at the output of the optical device; 27. The system of claim 26, further controlling the at least one processor to perform pre-processing on the waveform and convert the waveform into the eye diagram.

28. 26. The system of claim 25, wherein the trained neural network comprises a convolutional neural network.

29. The convolutional neural network includes:

30. The system of claim 28, comprising a plurality of extraction layers configured to perform convolution filtering, the optical output data being fed to the extraction layers with the input of a subsequent extraction layer being the output of a previous extraction layer.

30. The convolutional neural network includes: an input extraction layer comprising a convolution filter and a pooling layer, the input extraction layer being fed with the optical output data; a second extraction layer having a convolution filter and a pooling layer and receiving the output of the input extraction layer; a third extraction layer having a convolution filter and a pooling layer, the third extraction layer being fed with an output of the second extraction layer.

31. 26. The system of claim 25, wherein the trained neural network comprises a two-dimensional convolutional neural network.

32. 26. The system of claim 25, wherein the trained neural network comprises a one-dimensional convolutional neural network.

Images