Reduced false positive identification for spectral classification

Abstract

The present disclosure relates to reduced false positive identification for spectral classification. A device may receive information identifying a result of a set of spectral measurements of a training set of known samples and a validation set of known samples. The device may generate a classification model based on the information identifying the result of the set of spectral measurements, wherein the classification model includes at least one class associated with a material of interest for spectrometric determination, and wherein the classification model includes a mismatch class associated with at least one material of no interest or a baseline spectral measurement. The device may receive information identifying a particular result of a particular spectral measurement of an unknown sample. The device may determine whether the unknown sample is included in the mismatch class using the classification model. The device may provide an output indicating whether the unknown sample is included in the mismatch class.

Description

[0001] This application is a divisional application of an application filed on January 21, 2019, with application number 201910054785.3 and invention name “Reduced false positive identification for spectral classification”. Technical Field

[0002] The present application relates to, but is not limited to, reduced false positive identification for spectral classification. background

[0003] Raw material identification can be used for quality control of pharmaceutical products. For example, raw material identification can be performed on medical materials to determine whether the composition of the medical material corresponds to the packaging label associated with the medical material. Similarly, raw material quantification can be performed to determine the concentration of a specific chemical in a specific sample. Compared to other chemometric techniques, spectrometry can facilitate non-destructive identification and / or quantification of raw materials with reduced preparation and data acquisition time.

[0004] Overview

[0005] According to some possible embodiments, the device may include one or more memories and one or more processors communicatively coupled to the one or more memories. The device may receive information identifying the results of a set of spectral measurements of a training set of known samples and a validation set of known samples. The device may generate a classification model based on the information identifying the results of a set of spectral measurements, wherein the classification model includes at least one category associated with a material of interest for spectroscopic determination, and wherein the classification model includes a no-match class associated with at least one material of no interest or at least one of the baseline spectral measurements. The device may receive information identifying a specific result of a specific spectral measurement of an unknown sample. The device may use the classification model to determine whether the unknown sample is included in the no-match class. The device may provide an output indicating whether the unknown sample is included in the no-match class.

[0006] According to some possible implementations, a non-transitory computer-readable medium may store one or more instructions that, when executed by one or more processors, cause the one or more processors to receive information identifying a result of a spectral measurement performed on an unknown sample. When executed by the one or more processors, the one or more instructions may cause the one or more processors to aggregate multiple categories of a classification model to generate an aggregated classification model. When executed by the one or more processors, the one or more instructions may cause the one or more processors to use the aggregated classification model to determine that the spectral measurement was accurately performed. When executed by the one or more processors, the one or more instructions may cause the one or more processors to determine, based on determining that the spectral measurement was accurately performed, that the unknown sample is not included in a mismatch category of the classification model using the classification model, where the mismatch category is associated with a material of no interest or a baseline spectral measurement. When executed by the one or more processors, the one or more instructions may cause the one or more processors to perform spectral classification of the unknown sample based on determining that the unknown sample is not included in the mismatch category. When executed by the one or more processors, the one or more instructions may cause the one or more processors to provide information identifying the unknown sample based on performing spectral classification on the unknown sample.

[0007] According to some possible embodiments, the method may include obtaining, by a device, a result of a set of spectral measurements. The method may include generating, by the device, a classification model based on a support vector machine (SVM) based on the result of the set of spectral measurements, the classification model including a plurality of categories corresponding to a plurality of materials of interest for classification, wherein the set of spectral measurements includes measurements of a threshold number of samples of the plurality of materials of interest, wherein the classification model includes a specific category that does not correspond to the plurality of materials of interest for classification, and wherein the set of spectral measurements includes measurements of less than a threshold number of samples associated with the specific category. The method may include classifying, by the device, specific spectral measurements of specific samples into a specific category using the classification model. The method may include providing, by the device, information indicating that the specific sample is assigned to a specific category based on the classification of the specific spectral measurements.

[0008] 1) A device comprising:

[0009] one or more memories; and

[0010] one or more processors communicatively coupled to the one or more memories, the one or more processors configured to:

[0011] receiving information identifying results of a set of spectral measurements of a training set of known samples and a validation set of known samples;

[0012] generating a classification model based on the information identifying results of the set of spectral measurements, the classification model including at least one class associated with a material of interest for spectrometry, the classification model including a mismatch class associated with at least one of at least one material not of interest or a baseline spectral measurement;

[0013] receiving information identifying a specific result of a specific spectral measurement of an unknown sample;

[0014] Determining whether the unknown sample is included in the unmatched category using the classification model; and

[0015] An output is provided indicating whether the unknown sample is included in the no-match category.

[0016] 2) The apparatus according to 1), wherein the one or more processors, when determining whether the unknown sample is included in the unmatched category:

[0017] determining, based on the classification model, that the unknown sample is included in the unmatched category; and

[0018] wherein the one or more processors, when providing an output indicating whether the unknown sample is included in the unmatched category:

[0019] An output is provided indicating that the unknown sample is included in the no-match class.

[0020] 3) The apparatus according to 1), wherein the one or more processors, when determining whether the unknown sample is included in the unmatched category, are configured to:

[0021] determining, based on the classification model, that the unknown sample is not included in the unmatched category;

[0022] determining a classification of the unknown sample using the classification model and based on determining that the unknown sample is not included in the unmatched class; and

[0023] wherein the one or more processors, when providing an output indicating whether the unknown sample is included in the no-match category, are configured to:

[0024] An output is provided that identifies the classification of the unknown sample.

[0025] 4) The apparatus of 1), wherein the one or more processors, upon receiving the information identifying the results of the set of spectral measurements, are configured to:

[0026] receiving information identifying a set of baseline spectral measurements; and

[0027] Wherein, when generating the classification model, the one or more processors are configured to:

[0028] The mismatch class of the classification model is trained based on the set of baseline spectral measurements.

[0029] 5) The apparatus of 4), wherein the set of baseline spectral measurements is associated with at least one of:

[0030] Measurements performed with incorrect measurement distances,

[0031] Measurements performed with an incorrect measurement background,

[0032] Measurements performed with incorrect measurement lighting, or

[0033] A measurement performed in the absence of a sample.

[0034] 6) The apparatus of 1), wherein the one or more processors, upon receiving the information identifying the results of the set of spectral measurements, are configured to:

[0035] receiving information identifying the at least one material of no interest; and

[0036] Wherein, when generating the classification model, the one or more processors are configured to:

[0037] The mismatch class of the classification model is trained based on the information identifying the at least one material of no interest.

[0038] 7) The apparatus according to 1), wherein the one or more processors, when determining whether the unknown sample is included in the mismatch category using the classification model:

[0039] A support vector machine-based confidence metric is used to determine whether the unknown sample is included in the unmatched category.

[0040] 8) The apparatus according to 7), wherein the confidence metric is at least one of the following:

[0041] probability estimates, or

[0042] Decision value.

[0043] 9) The apparatus according to 1), wherein the classification model is a first classification model; and

[0044] wherein, when determining whether the unknown sample is included in the unmatched category, the one or more processors:

[0045] performing a first classification using the first classification model to identify a set of local classes of the first classification model for the particular spectral measurement;

[0046] generating a second classification model based on the set of local categories, the second classification model including the mismatched category; and

[0047] A second classification is performed to determine whether the unknown sample is included in the unmatched class.

[0048] 10) A non-transitory computer-readable medium storing instructions, the instructions comprising:

[0049] One or more instructions that, when executed by one or more processors, cause the one or more processors to:

[0050] receiving information identifying results of a spectroscopic measurement performed on an unknown sample;

[0051] aggregating multiple categories of the classification model to generate an aggregated classification model;

[0052] determining, using the aggregated classification model, that the spectral measurements were accurately performed;

[0053] based on determining that the spectral measurement was accurately performed and using the classification model, determining that the unknown sample is not included in a mismatch category of the classification model, the mismatch category being associated with material not of interest or a baseline spectral measurement;

[0054] performing spectral classification of the unknown sample based on determining that the unknown sample is not included in the unmatched category; and

[0055] Based on performing the spectral classification of the unknown sample, information identifying the unknown sample is provided.

[0056] 11) The non-transitory computer-readable medium of 10), wherein the one or more instructions causing the one or more processors to determine that the unknown sample is not included in the mismatch category cause the one or more processors to:

[0057] Based on a confidence metric associated with the classification model satisfying a threshold, it is determined that the unknown sample is not included in the unmatched category.

[0058] 12) The non-transitory computer-readable medium of 11), wherein the one or more instructions, when executed by the one or more processors, further cause the one or more processors to:

[0059] The confidence metric is determined based on partitioning the classification model into a plurality of sub-models using a one-vs-all technique or an all-pairs technique.

[0060] 13) The non-transitory computer-readable medium according to 11), wherein the classification model includes more than a threshold number of categories; and

[0061] wherein the one or more instructions causing the one or more processors to perform the spectral classification cause the one or more processors to:

[0062] performing a first spectral classification of the unknown sample based on the classification model;

[0063] generating another classification model based on performing the first spectral classification using a subset of categories of the classification model;

[0064] Determining based on the other classification model that the unknown sample is not included in the unmatched category; and

[0065] A second classification is performed to identify the unknown sample.

[0066] 14) A method comprising:

[0067] A set of spectral measurement results are obtained by the device;

[0068] generating, by the device, a support vector machine (SVM)-based classification model based on results of the set of spectral measurements, the classification model including a plurality of classes corresponding to a plurality of materials of interest for classification, the set of spectral measurements including a threshold number of measurements of samples of the plurality of materials of interest, the classification model including a particular class that does not correspond to the plurality of materials of interest for classification, the set of spectral measurements including less than the threshold number of measurements of samples associated with the particular class;

[0069] classifying, by the device, a particular spectral measurement of a particular sample into the particular category using the classification model; and

[0070] Information indicating that the particular sample is assigned to the particular category is provided by the device based on classifying the particular spectral measurement.

[0071] 15) The method according to 14), wherein classifying the specific spectral measurement comprises:

[0072] dividing the classification model into a plurality of sub-models, each sub-model in the plurality of sub-models corresponding to a comparison between a corresponding category of the classification model and each other category of the classification model;

[0073] determining a plurality of decision values ​​corresponding to the plurality of sub-models; and

[0074] The specific category is selected for the specific sample based on the multiple decision values.

[0075] 16) The method according to 14), wherein classifying the specific spectral measurement comprises:

[0076] dividing the classification model into a plurality of sub-models, the plurality of sub-models corresponding to comparisons between each category of the classification model;

[0077] determining a plurality of decision values ​​corresponding to the plurality of sub-models; and

[0078] The specific category is selected for the specific sample based on the multiple decision values.

[0079] 17) The method according to 14), wherein classifying the specific spectral measurement comprises:

[0080] The specific spectral measurement is classified using a kernel function of the radial basis function type or a kernel function of the linear kernel type.

[0081] 18) The method according to 14), wherein the set of spectral measurements includes a baseline spectral measurement and a spectral measurement of a material not of interest; and

[0082] The baseline spectral measurements and the spectral measurements of the materials not of interest are classified into the particular categories.

[0083] 19) The method according to 14), wherein classifying the specific spectral measurement comprises:

[0084] The specific spectral measurement is classified using an in-situ local classification model generated based on the classification model.

[0085] 20) The method according to 14), wherein classifying the specific spectral measurement comprises:

[0086] Aggregating the categories of the classification model into a single category; and

[0087] The particular spectral measurement is classified based on the single category. BRIEF DESCRIPTION OF THE DRAWINGS

[0088] Figure 1A and Figure 1B is a diagram that provides an overview of example implementations described herein;

[0089] Figure 2 is a diagram of an example environment in which the systems and / or methods described herein may be implemented;

[0090] Figure 3yes Figure 2 diagrams of example components of one or more devices;

[0091] Figure 4 is a flow chart of an example process for generating a classification model for spectral classification;

[0092] Figure 5 is with Figure 4 Figures of example implementations related to the example processes shown in ;

[0093] Figure 6 is a flow chart of an example process for avoiding false positive identifications during spectral classification; and

[0094] Figure 7A and Figure 7B is with Figure 6 FIGURES of example implementations related to the example processes shown in FIG. Detailed description

[0095] The following detailed description of example embodiments refers to the accompanying drawings, in which the same reference numbers in different drawings may identify the same or similar elements.

[0096] Raw material identification (RMID) is a technique used to identify the components (e.g., ingredients) of a particular sample for identification, verification, etc. For example, RMID can be used to verify that the components in a pharmaceutical material correspond to the set of ingredients identified on the label. Similarly, raw material quantification is a technique used to perform quantitative analysis on a particular sample, such as determining the concentration of a particular material in a particular sample. A spectrometer can be used to perform spectroscopy on a sample (e.g., a pharmaceutical material) to determine the components of the sample, the concentrations of the components of the sample, etc. The spectrometer can determine a set of measurements of the sample and can provide a set of measurements for spectrometric determination. Spectral classification techniques (e.g., classifiers) can help determine the components of a sample based on a set of measurements of the sample.

[0097] However, some unknown samples that are to be spectrally classified are not actually included in the categories that the classification model is configured to classify. For example, a user may inadvertently provide beef for classification in a classification model trained to distinguish between types of fish. In this case, the control device may perform spectral classification of a specific material and may provide a false positive identification of the specific material as a specific type of fish, which would be inaccurate.

[0098] As another example, a classification model can be trained to classify types of sugars (e.g., glucose, fructose, galactose, etc.) and quantify the respective concentrations of each type of sugar in an unknown sample. However, a user of the spectrometer and control device may inadvertently attempt to classify an unknown sample of sugars based on improper use of the spectrometer to perform measurements. For example, the user may operate the spectrometer at an incorrect distance from the unknown sample, under environmental conditions that differ from calibration conditions, and / or similar conditions under which spectroscopy is performed to train the classification model. In this case, the control device may receive an inaccurate spectrum for the unknown sample when the unknown sample is actually a second type of sugar at a second concentration, resulting in a false positive identification of the unknown sample as a first type of sugar at a first concentration.

[0099] Some embodiments described herein can utilize mismatch categories for classification models to reduce false positive identifications for spectroscopy. For example, a control device that receives a spectral measurement of an unknown sample can determine whether to assign the unknown sample to a mismatch category. In some embodiments, the control device can determine that the unknown sample is to be assigned to a mismatch category and can provide information indicating that the unknown sample is assigned to the mismatch category, thereby avoiding false positive identifications of the unknown sample. Alternatively, based on determining that the unknown sample is not assigned to the mismatch category, the control device can analyze the spectrum of the unknown sample to provide, for example, a spectral determination of classification, concentration, etc. In addition, the control device can utilize a confidence metric (such as a probability estimate, a decision value, etc.) to filter out false positive identifications.

[0100] In this manner, the accuracy of the spectroscopy method is improved relative to a spectroscopy method performed without using a mismatch class and / or confidence metric. Furthermore, the mismatch class can be used when generating a classification model based on a training set of known spectral samples. For example, the control device can determine that a sample of the training set is not of a type that corresponds to the rest of the training set (e.g., due to human error that resulted in an erroneous sample being introduced into the training set), and can determine not to include data regarding the sample when generating the classification model. In this manner, the control device improves the accuracy of the classification model used for the spectroscopy method.

[0101] Figure 1A and Figure 1B is a diagram of an overview of an example implementation 100 described herein. Figure 1A As shown, example embodiment 100 may include a control device and a spectrometer.

[0102] like Figure 1AAs further shown in , the control device can cause the spectrometer to perform a set of spectral measurements on a training set and a validation set (e.g., a known sample set used for training and validation of a classification model). The training set and the validation set can be selected to include a threshold number of samples for each category of the classification model. The categories of the classification model can refer to groupings of similar materials that share one or more common characteristics, such as (in the context of pharmaceuticals) lactose materials, fructose materials, acetaminophen materials, ibuprofen materials, aspirin materials, etc. The materials used to train the classification model and for which the classification model is to be used to perform raw material identification can be referred to as materials of interest.

[0103] like Figure 1A As further shown in FIG, the spectrometer can perform a set of spectral measurements on the training set and the validation set based on instructions received from the control device. For example, the spectrometer can determine the spectrum of each sample in the training set and the validation set to enable the control device to generate a set of classes for classifying an unknown sample as one of the materials of interest for the classification model.

[0104] The spectrometer can provide a set of spectral measurements to the control device. The control device can use a specific measurement technique and generate a classification model based on the set of spectral measurements. For example, the control device can use support vector machine (SVM) technology (e.g., a machine learning technique for information measurement) to generate a global classification model. The global classification model can include information associated with assigning a specific spectrum to a specific class of material of interest, and can include information associated with identifying the type of material of interest associated with the specific class. In this way, the control device can provide information identifying the type of material of an unknown sample based on assigning the spectrum of the unknown sample to a specific class.

[0105] In some embodiments, the control device may receive spectra associated with samples for a mismatch category. For example, the control device may receive spectra determined to be similar to the spectrum of a material of interest, spectra associated with materials that may be confused with the material of interest (e.g., visually, chemically, etc.), spectra associated with incorrect operation of the spectrometer (e.g., spectra measured without a sample, spectra measured at an incorrect distance between the sample and the optics of the spectrometer, etc.), and / or similar spectra. Materials that are not materials of interest and that may be included in the mismatch category may be referred to as nuisance materials or materials of no interest. In this case, the control device may generate a mismatch category for the classification model and may use the mismatch category to verify false positive identification avoidance based on the spectra of the nuisance materials included in the validation set. Additionally or alternatively, during use of the classification model, the control device may receive information identifying the nuisance material and may update the classification model to be able to avoid false positive identifications (e.g., identifying the nuisance material as one of the materials of interest).

[0106] like Figure 1B As shown, a control device may receive a classification model (e.g., from a memory, from another control device that generates a classification model, etc.). The control device may cause a spectrometer to perform a set of spectral measurements on an unknown sample (e.g., an unknown sample for which classification or quantification is to be performed). The spectrometer may perform the set of spectral measurements based on instructions received from the control device. For example, the spectrometer may determine a spectrum for the unknown sample. The spectrometer may provide the set of spectral measurements to the control device. The control device may attempt to classify the unknown sample based on the classification model, for example, using a multi-stage classification technique.

[0107] about Figure 1B , the control device may attempt to use the classification model to determine whether the unknown sample is in the mismatch class. For example, the control device may determine a confidence metric corresponding to the likelihood that the unknown sample belongs to the mismatch class. In this case, based on the control device's determination that the confidence metric (e.g., a probability estimate, a decision value output of a support vector machine, etc.) satisfies a threshold, the control device may assign the unknown sample to the mismatch class. In this case, the control device may report that the unknown sample cannot be accurately classified using the classification model, thereby reducing the likelihood that the unknown sample will be falsely identified as belonging to the class of the material of interest.

[0108] In some embodiments, based on a first determination that an unknown sample does not belong to an unmatched category, the control device may attempt to perform a determination of a specific sample of the unknown set using in-situ local modeling. For example, the control device may determine a set of confidence metrics associated with the specific sample and the global classification model. In this case, the control device may select a subset of categories of the global classification model based on one or more corresponding confidence metrics, and may generate a local classification model based on the set of categories. The local classification model may be an in-situ classification model generated using SVM technology and a subset of categories. Based on generating the in-situ classification model, the control device may attempt to classify the unknown sample based on the local classification model. In this case, based on one or more confidence metrics associated with the local classification model meeting a threshold, the control device may determine that the unknown sample does belong to the unmatched category, and may report that the unknown sample cannot be classified using the classification model. Alternatively, the control device may determine that the unknown sample does not belong to the unmatched category, and may report a classification associated with the unknown sample.

[0109] In this manner, the control device implements spectrometry for unknown samples with improved accuracy relative to other classification models based on reducing the likelihood of false positive identifications of reporting unknown samples as materials of interest.

[0110] As indicated above, Figure 1A and Figure 1B It is provided as an example only. Other examples are possible and may differ from the Figure 1A and Figure 1B An example of description.

[0111] Figure 2 is a diagram of an example environment 200 in which the systems and / or methods described herein may be implemented. Figure 2 As shown, environment 200 may include a control device 210, a spectrometer 220, and a network 230. The devices of environment 200 may be interconnected via wired connections, wireless connections, or a combination of wired and wireless connections.

[0112] The control device 210 may include one or more devices capable of storing, processing, and / or routing information associated with spectral classification. For example, the control device 210 may include a server, a computer, a wearable device, a cloud computing device, and / or a similar device that generates a classification model based on a set of measurements from a training set, validates the classification model based on a set of measurements from a validation set, and / or utilizes the classification model to perform spectral classification based on a set of measurements from an unknown set. In some embodiments, as described herein, the control device 210 may utilize machine learning techniques to determine whether the spectral measurements of an unknown sample will be classified into a mismatch category to reduce the likelihood of false positive identifications. In some embodiments, the control device 210 may be associated with a specific spectrometer 220. In some embodiments, the control device 210 may be associated with multiple spectrometers 220. In some embodiments, the control device 210 may receive information from another device in the environment 200 (e.g., a spectrometer 220) and / or transmit information to another device in the environment 200 (e.g., a spectrometer 220).

[0113] Spectrometer 220 may include one or more devices capable of performing spectroscopic measurements on a sample. For example, spectrometer 220 may include a spectrometer device that performs spectroscopy (e.g., vibrational spectroscopy, such as near-infrared (NIR) spectroscopy, mid-infrared spectroscopy (mid-IR), Raman spectroscopy, etc.). In some embodiments, spectrometer 220 may be incorporated into a wearable device, such as a wearable spectrometer and / or the like. In some embodiments, spectrometer 220 may receive information from another device in environment 200 (e.g., control device 210) and / or transmit information to another device in environment 200 (e.g., control device 210).

[0114] The network 230 may include one or more wired networks and / or wireless networks. For example, the network 230 may include a cellular network (e.g., a Long Term Evolution (LTE) network, a 3G network, a Code Division Multiple Access (CDMA) network, etc.), a public land mobile network (PLMN), a local area network (LAN), a wide area network (WAN), a metropolitan area network (MAN), a telephone network (e.g., a public switched telephone network (PSTN)), a private network, an ad hoc network, an intranet, the Internet, a fiber-optic-based network, a cloud computing network, etc., and / or a combination of these or other types of networks.

[0115] Figure 2 The number and arrangement of devices and networks shown in the FIGURES are provided as examples. Figure 2 There may be additional devices and / or networks, fewer devices and / or networks, different devices and / or networks, or differently arranged devices and / or networks than those shown in FIG. Figure 2Two or more devices shown in FIG may be implemented in a single device, or Figure 2 A single device shown in FIG200 may be implemented as multiple distributed devices. For example, although control device 210 and spectrometer 220 are described herein as two separate devices, control device 210 and spectrometer 220 may be implemented within a single device. Additionally or alternatively, one or more devices of environment 200 (e.g., one or more devices) may perform one or more functions described as being performed by another group of devices of environment 200.

[0116] Figure 3 is a diagram of example components of device 300. Device 300 may correspond to control device 210 and / or spectrometer 220. In some embodiments, control device 210 and / or spectrometer 220 may include one or more devices 300 and / or one or more components of device 300. Figure 3 As shown, device 300 may include a bus 310 , a processor 320 , a memory 330 , a storage component 340 , an input component 350 , an output component 360 , and a communication interface 370 .

[0117] The bus 310 includes components that allow communication among the components of the device 300. The processor 320 is implemented in hardware, firmware, or a combination of hardware and software. The processor 320 is a central processing unit (CPU), a graphics processing unit (GPU), an accelerated processing unit (APU), a microprocessor, a microcontroller, a digital signal processor (DSP), a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), or other types of processing components. In some embodiments, the processor 320 includes one or more processors that can be programmed to perform functions. The memory 330 includes a random access memory (RAM), a read-only memory (ROM), and / or another type of dynamic or static storage device (e.g., flash memory, magnetic memory, and / or optical memory) that stores information and / or instructions for use by the processor 320.

[0118] The storage component 340 stores information and / or software related to the operation and use of the device 300. For example, the storage component 340 may include a hard disk (e.g., a magnetic disk, an optical disk, a magneto-optical disk, and / or a solid-state disk), a compact disk (CD), a digital versatile disk (DVD), a floppy disk, a cartridge, a magnetic tape, and / or another type of non-transitory computer-readable medium along with a corresponding drive.

[0119] Input components 350 include components that allow device 300 to receive information, for example, via user input (e.g., a touch screen display, a keyboard, a keypad, a mouse, buttons, switches, and / or a microphone). Additionally or alternatively, input components 350 may include sensors for sensing information (e.g., a global positioning system (GPS) component, an accelerometer, a gyroscope, and / or an actuator). Output components 360 include components that provide output information from device 300 (e.g., a display, a speaker, and / or one or more light emitting diodes (LEDs)).

[0120] The communication interface 370 includes transceiver-like components (e.g., a transceiver and / or a separate receiver and transmitter) that enable the device 300 to communicate with other devices, for example, via a wired connection, a wireless connection, or a combination of wired and wireless connections. The communication interface 370 can allow the device 300 to receive information from another device and / or provide information to another device. For example, the communication interface 370 can include an Ethernet interface, an optical interface, a coaxial interface, an infrared interface, a radio frequency (RF) interface, a universal serial bus (USB) interface, a wireless local area network interface, a cellular network interface, etc.

[0121] Device 300 can perform one or more processes described herein. Device 300 can perform these processes based on processor 320 executing software instructions stored by non-transitory computer-readable media (e.g., memory 330 and / or storage component 340). Computer-readable media is defined herein as non-transitory memory devices. Memory devices include memory space within a single physical storage device or memory space spread across multiple physical storage devices.

[0122] The software instructions may be read into the memory 330 and / or storage component 340 from another computer-readable medium or from another device via the communication interface 370. The software instructions stored in the memory 330 and / or storage component 340, when executed, may cause the processor 320 to perform one or more processes described herein. Additionally or alternatively, hardwired circuitry may be used in place of or in combination with software instructions to perform one or more processes described herein. Thus, the embodiments described herein are not limited to any specific combination of hardware circuitry and software.

[0123] Figure 3 The number and arrangement of components shown in FIG are provided as examples. In practice, Figure 3 300 may include additional components, fewer components, different components, or differently arranged components than those shown in FIG. Additionally or alternatively, one or more components of device 300 may perform one or more functions described as being performed by another set of components of device 300.

[0124] Figure 4 is a flow chart of an example process 400 for generating a classification model for spectral classification. In some embodiments, Figure 4 One or more process blocks of may be performed by the control device 210. In some implementations, Figure 4 One or more process blocks of may be performed by another device or group of devices separate from or including the control device 210 , such as the spectrometer 220 .

[0125] like Figure 4 As shown, process 400 may include causing a set of spectral measurements to be performed on a training set and / or a validation set (block 410). For example, control device 210 may (e.g., using processor 320, communication interface 370, etc.) cause spectrometer 220 to perform a set of spectral measurements on a training set and / or a validation set of samples to determine a spectrum for each sample in the training set and / or the validation set. A training set may refer to a set of samples of one or more known materials that are used to generate a classification model. Similarly, a validation set may refer to a set of samples of one or more known materials that are used to verify the accuracy of a classification model. For example, a training set and / or a validation set may include one or more versions of a set of materials (e.g., one or more versions manufactured by different manufacturers to control for manufacturing variations).

[0126] In some embodiments, the training set and / or validation set can be selected based on an expected set of materials of interest for which spectral classification will be performed using the classification model. For example, when spectral quantification is expected to be performed on a pharmaceutical material to determine the presence of a specific component of the pharmaceutical material, the training set and / or validation set can include a sample set of active pharmaceutical ingredients (APIs), excipients, etc. at a set of different possible concentrations.

[0127] In some embodiments, the training set and / or validation set can be selected to include a specific number of samples of each type of material. For example, the training set and / or validation set can be selected to include multiple samples (e.g., 5 samples, 10 samples, 15 samples, 50 samples, etc.) of a specific material and / or its concentration. In some embodiments, the number of samples can be less than a threshold value. For example, categories of homogeneous organic compounds can be generated based on 50 spectra (e.g., spectral scans) of 10 samples, 15 spectra of 3 samples, etc. Similarly, categories of heterogeneous organic compounds can be generated based on, for example, 100 spectra from 20 samples, 50 spectra from 10 samples, etc. Similarly, categories of biological or agricultural materials can be generated based on 400 spectra from 40 samples, 200 spectra from 20 samples, etc. In some embodiments, the number of samples and / or spectra for a no-match class of interfering materials can be associated with the same or a reduced number of samples and / or spectra for a non-no-match class of the same type of material (e.g., homogeneous organic compounds, heterogeneous organic compounds, biological or agricultural materials, etc.). In this manner, the control device 210 can be provided with a threshold number of spectra associated with a particular type of material, thereby facilitating the generation and / or validation of a classification model (e.g., a global classification model, a local classification model, etc.) or a quantitative model to which an unknown sample can be accurately assigned, which can be used to quantify spectra assigned to a class associated with the quantitative model.

[0128] In some embodiments, one or more samples of materials to be assigned to the mismatch category may be included in the training set and / or validation set. For example, the spectrometer 220 may provide measurements of a first material that is associated with a spectrum similar to a second material to be quantified using the quantization model. In this way, the control device 210 may use machine learning to train the avoidance of false positive identification. In some embodiments, the control device 210 may select materials for the mismatch category based on the received information. For example, the control device 210 may receive information identifying interfering materials having a spectrum, appearance, etc. similar to a material of interest at a specific concentration for which a classification model is to be generated. Additionally or alternatively, the control device 210 may perform machine learning techniques to automatically identify interfering materials for a specific material of interest. For example, the control device 210 may use machine learning to perform pattern recognition to identify spectra of interfering materials that are similar to the spectrum of the material of interest, to identify interfering materials that appear visually similar to the material of interest, and / or the like.

[0129] In some embodiments, the control device 210 may cause a baseline spectral measurement to be performed to identify spectra that do not match the classification. For example, the control device 210 may cause a spectral measurement to be performed without a sample, with an incorrect background, with incorrect lighting, and / or the like as a baseline spectral measurement to ensure that incorrect spectral measurements are classified as a mismatch rather than as belonging to a particular material of interest. In this case, the control device 210 may automatically control the spectrometer 220, provide information using a user interface to guide the user of the spectrometer 220 in performing the incorrect measurement, and / or the like. Additionally or alternatively, the control device 210 may receive information indicating that a particular spectral measurement was incorrectly performed, resulting in the generation of a mismatch classification.

[0130] In some embodiments, the control device 210 can cause multiple spectrometers 220 to perform a set of spectral measurements to address one or more physical conditions. For example, the control device 210 can cause the first spectrometer 220 and the second spectrometer 220 to perform a set of vibrational spectral measurements using NIR spectroscopy. Additionally or alternatively, the control device 210 can cause the set of spectral measurements to be performed at multiple times, at multiple locations, under multiple different laboratory conditions, and so on. In this manner, the control device 210 reduces the likelihood of inaccurate spectral measurements as a result of the physical conditions relative to causing a set of spectral measurements to be performed by a single spectrometer 220.

[0131] like Figure 4 As further shown in FIG4 , process 400 may include receiving information identifying the results of a set of spectral measurements (block 420). For example, control device 210 may receive information identifying the results of the set of spectral measurements (e.g., using processor 320, communication interface 370, etc.). In some embodiments, control device 210 may receive information identifying a set of spectra corresponding to samples in a training set and / or a validation set. For example, control device 210 may receive information identifying specific spectra observed when spectrometer 220 performed spectroscopy on the training set. In some embodiments, control device 210 may receive information identifying spectra for both the training set and the validation set. In some embodiments, control device 210 may receive information identifying the spectra for the training set, generate a classification model, and, after generating the classification model, receive information identifying the spectra for the validation set to test the classification model. In some embodiments, control device 210 may receive other information as a result of the set of spectral measurements, such as information indicating that the measurements were performed inaccurately, resulting in a mismatched class. Additionally or alternatively, control device 210 may receive information associated with identifying energy absorption, energy radiation, energy scattering, etc.

[0132] In some embodiments, the control device 210 can receive information identifying results of a set of spectral measurements from the plurality of spectrometers 220. For example, the control device 210 can control physical conditions (e.g., differences between the plurality of spectrometers 220, potential differences in laboratory conditions, etc.) by receiving spectral measurements performed by the plurality of spectrometers 220, performed at a plurality of different times, performed at a plurality of different locations, etc.

[0133] In some embodiments, the control device 210 can remove one or more spectra from use in generating a classification model. For example, the control device 210 can perform spectral classification and classify a spectrum into a mismatch category, and can determine that the sample corresponding to the spectrum is unintentionally an interfering material or a material of no interest (e.g., based on human error in correctly performing the spectroscopy, errors in the information identifying the spectra in the training set, etc.), and can determine to remove the spectrum from the training set. In this way, the control device 210 can improve the accuracy of the classification model by reducing the likelihood that incorrect or inaccurate information about the training set or validation set will be used to generate the classification model.

[0134] like Figure 4 As further shown in FIG4 , process 400 may include generating a classification model based on information identifying the results of a set of spectral measurements (block 430). For example, control device 210 may generate a global classification model associated with a principal component analysis (PCA)-SVM classifier technique (e.g., for use in an in-situ local modeling technique) based on information identifying the results of a set of spectral measurements (e.g., using processor 320, memory 330, storage component 340, etc.).

[0135] In some embodiments, the control device 210 may perform a set of measurements to generate a global classification model. For example, the control device 210 may generate a set of categories for the global classification model and may assign a set of spectra identified from the results of a set of spectral measurements to local categories based on the use of support vector machines (SVMs). In some embodiments, while using the global classification model, the control device 210 uses a confidence metric associated with the global classification model to identify a threshold number of local categories corresponding to the unknown spectrum, generates a local classification model based on the local categories, and determines the identity of the unknown spectrum based on the local classification model. In this case, mismatched categories may be generated for the local classification model (e.g., the local classification model generated in situ from the global classification model may include mismatched categories). In this manner, by using in situ local modeling with a first classification and a second classification, the control device 210 is able to classify a large number of categories (e.g., greater than a threshold, such as greater than 50 categories, greater than 100 categories, greater than 200 categories, greater than 1000 categories, etc.). In some embodiments, the control device 210 may generate another type of classification model for classifying the unknown spectrum and / or use another type of classifier for the classification model.

[0136] SVM can refer to a supervised learning model that performs pattern recognition and uses a confidence metric for classification. In some embodiments, when using SVM technology to generate a global classification model, the control device 210 can use a specific type of kernel function to determine the similarity of two or more inputs (e.g., spectra). For example, the control device 210 can use a kernel function of the radial basis function (RBF) (e.g., referred to as SVM-rbf) type, which can be expressed as k(x,y)=exp(-||xy||^2) for spectra x and y; a kernel function of the linear function (e.g., when used in a multi-stage determination technique, referred to as SVM-linear and hierarchical (hier)-SVM-linear) type, which can be expressed as k(x,y)=<x·y> ; a kernel function of the sigmoid function type; a kernel function of the polynomial function type; a kernel function of the exponential function type; and / or similar functions.

[0137] In some embodiments, the control device 210 can utilize a specific type of confidence metric for the SVM, such as a probability-based SVM (e.g., a measure based on determining the probability that a sample is a member of a class from a set of classes), a decision-based SVM (e.g., a measure that utilizes a decision function to vote for a class from a set of classes as the class of which the sample is a member), etc. For example, during use of a classification model utilizing a decision-based SVM, the control device 210 can determine whether the unknown sample is within the boundaries of the constituent classes based on mapping the spectrum of the unknown sample, and can assign the sample to a class based on whether the unknown sample is within the boundaries of the constituent classes. In this manner, the control device 210 can determine whether to assign the unknown spectrum to a particular class, to a mismatch class, etc.

[0138] In some embodiments, the control device 210 can utilize a specific class comparison technique to determine the decision value. For example, the control device 210 can utilize a one-versus-all technique (sometimes referred to as a one-versus-all technique), in which the classification model is divided into a set of sub-models, each sub-model is based on a comparison of one class with every other class of the classification model, and the decision value is determined based on each sub-model. Additionally or alternatively, the control device 210 can utilize an all-pair technique, in which the classification model is divided into every possible class pair to form a sub-model, and the decision value is determined from the sub-model.

[0139] Although some embodiments described herein are described in terms of a particular set of machine learning techniques, other techniques may also be used to determine information about unknown spectra, such as classification of materials, etc.

[0140] In some embodiments, the control device 210 can select a particular classifier from a set of classification techniques to be used to generate the global classification model. For example, the control device 210 can generate multiple classification models corresponding to multiple classifiers, and can test the multiple classification models, for example by determining each model's transferability (e.g., how accurate the classification model generated based on spectral measurements performed on a first spectrometer 220 is when applied to spectral measurements performed on a second spectrometer 220), large-scale measurement accuracy (e.g., how accurately the classification model can be used to simultaneously classify a certain number of samples that meet a threshold), etc. In this case, the control device 210 can select a classifier, such as an SVM classifier (e.g., hierarchical-SVM-linear), based on determining that the classifier is associated with superior transferability and / or large-scale measurement accuracy relative to other classifiers.

[0141] In some embodiments, the control device 210 can generate a classification model based on information identifying samples of the training set. For example, the control device 210 can use information identifying the type or concentration of the material represented by the samples of the training set to identify the class of spectra having the type or concentration of the material. In some embodiments, when generating the classification model, the control device 210 can train the classification model. For example, the control device 210 can use a portion of a set of spectral measurements (e.g., measurements associated with the training set) to train the model. Additionally or alternatively, the control device 210 can perform an evaluation of the classification model. For example, the control device 210 can validate the classification model (e.g., for predicted strength) using another portion of the set of spectral measurements (e.g., a validation set).

[0142] In some embodiments, the control device 210 can use a multi-stage determination technique to validate the classification model. For example, for classification based on in-situ local modeling, the control device 210 can determine that the global classification model is accurate when utilized in conjunction with one or more local classification models. In this manner, the control device 210 ensures that the classification model is generated with a threshold accuracy before providing it for utilization, for example, by the control device 210, by other control devices 210 associated with other spectrometers 220, and the like.

[0143] In some embodiments, the control device 210 may provide the classification model to other control devices 210 associated with other spectrometers 220 after generating the classification model. For example, a first control device 210 may generate a classification model and may provide the classification model to a second control device 210 for utilization. In this case, for classification based on in-situ local modeling, the second control device 210 may store the classification model (e.g., a global classification model) and may utilize the classification model to generate one or more in-situ local classification models for classifying one or more samples of an unknown set. Additionally or alternatively, the control device 210 may store the classification model for utilization by the control device 210 when performing classification, generating one or more local classification models (e.g., for classification based on in-situ local modeling), and / or the like. In this manner, the control device 210 provides the classification model for utilization in spectral classification of unknown samples.

[0144] Although Figure 4 Example blocks of process 400 are shown, but in some implementations, Figure 4 Process 400 may include additional blocks, fewer blocks, different blocks, or differently arranged blocks compared to those depicted in . Additionally or alternatively, two or more blocks of process 400 may be executed in parallel.

[0145] Figure 5 is with Figure 4FIG. 4 is a diagram of an example implementation 500 related to the example process 400 shown in FIG. Figure 5 An example of generating a classification model with false positive identification for quantification is shown.

[0146] like Figure 5 As shown, the control device 210-1 transmits information to the spectrometer 220-1 to instruct the spectrometer 220-1 to perform a set of spectral measurements on the training set and validation set 510. Assume that the training set and validation set 510 include a first set of training samples (e.g., whose measurements are used to train the classification model) and a second set of validation samples (e.g., whose measurements are used to verify the accuracy of the classification model). As shown by reference numeral 515, the spectrometer 220-1 performs a set of spectral measurements based on the received instructions. As shown by reference numeral 520, the control device 210-1 receives the first set of spectra for the training samples and the second set of spectra for the validation samples. In this case, the validation samples may include samples of multiple materials of interest for classification and one or more interfering materials or one or more samples of incorrect measurements that do not match the classification model to avoid false positive identification. Assume that the control device 210-1 stores information identifying each sample of the training set and validation set 510.

[0147] about Figure 5 , assuming that the control device 210-1 has chosen to generate a classification model using a hierarchical-SVM-linear classifier (e.g., based on testing the hierarchical-SVM-linear classifier against one or more other classifiers), the classification model can be an in-situ local modeling type of classification model. As shown by reference numeral 525, the control device 210-1 trains the classification model using the hierarchical-SVM-linear classifier and the first set of spectra, and validates the classification model using the hierarchical-SVM-linear classifier and the second set of spectra. The control device 210-1 can use a subset of the first set of spectra to train the classification model to identify interfering materials, and use a subset of the second set of spectra to validate the accuracy of the classification model in identifying interfering materials, thereby generating a mismatch category for the classification model.

[0148] Assume that the control device 210-1 determines that the classification model satisfies the validation threshold (e.g., has an accuracy that exceeds the validation threshold). As shown by reference numeral 530, the control device 210-1 provides the classification model to the control device 210-2 (e.g., for use in classifying spectral measurements performed by the spectrometer 220-2) and to the control device 210-3 (e.g., for use in classifying spectral measurements performed by the spectrometer 220-3).

[0149] As indicated above, Figure 5 It is provided as an example only. Other examples are possible and may differ from the Figure 5 An example of description.

[0150] In this manner, the control device 210 facilitates the generation of a classification model based on a selected classification technique (e.g., a technique selected based on model transferability, accuracy of large-scale classification, etc.) and the distribution of the classification model for utilization by one or more other control devices 210 associated with one or more spectrometers 220. Additionally, the control device 210 improves the accuracy of the classification model by including spectral measurements of interfering materials to avoid false positive identifications.

[0151] Figure 6 is a flow chart of an example process 600 for avoiding false positive identifications during raw material identification. In some embodiments, Figure 6 One or more process blocks of may be performed by the control device 210. In some embodiments, Figure 6 One or more process blocks of may be performed by another device or group of devices separate from or including the control device 210 , such as the spectrometer 220 .

[0152] like Figure 6 As shown, process 600 may include receiving information identifying the results of a set of spectral measurements performed on an unknown sample (block 610). For example, control device 210 may receive (e.g., using processor 320, communication interface 370, etc.) information identifying the results of a set of spectral measurements performed on the unknown sample. In some embodiments, control device 210 may receive information identifying the results of a set of spectral measurements on an unknown set (e.g., a plurality of samples). The unknown set may include a set of samples (e.g., unknown samples) for which an assay (e.g., spectral classification) is to be performed. For example, control device 210 may cause spectrometer 220 to perform a set of spectral measurements on the set of unknown samples and may receive information identifying a set of spectra corresponding to the set of unknown samples.

[0153] In some embodiments, the control device 210 can receive information identifying results from multiple spectrometers 220. For example, the control device 210 can cause multiple spectrometers 220 to perform a set of spectral measurements on an unknown set (e.g., the same set of samples) and can receive information identifying a set of spectra corresponding to samples of the unknown set. Additionally or alternatively, the control device 210 can receive information identifying the results of a set of spectral measurements performed at multiple times, at multiple locations, etc., and can classify and / or quantify a particular sample based on the set of spectral measurements performed at multiple times, at multiple locations, etc. (e.g., based on averaging the set of spectral measurements or based on another technique). In this manner, the control device 210 can address physical conditions that may affect the results of the set of spectral measurements.

[0154] Additionally or alternatively, the control device 210 can cause the first spectrometer 220 to perform a first portion of a set of spectral measurements on a first portion of the unknown set, and can cause the second spectrometer 220 to perform a second portion of the set of spectral measurements on a second portion of the unknown set. In this manner, the control device 210 can reduce an amount of time to perform a set of spectral measurements relative to having all spectral measurements performed by a single spectrometer 220.

[0155] like Figure 6 As further shown in FIG6 , process 600 may include determining whether a set of spectral measurements were accurately performed (block 620). For example, control device 210 may determine (e.g., using processor 320, memory 330, storage component 340, etc.) whether the set of spectral measurements were accurately performed. In some embodiments, control device 210 may determine whether the spectral measurements of the unknown sample were performed at a calibrated distance (e.g., between the optical components of spectrometer 220 and the sample, between the optical components of spectrometer 220 and a background of the sample, etc.). Additionally or alternatively, control device 210 may determine whether the spectral measurements of the unknown sample were performed at a calibrated temperature, at a calibrated pressure, at a calibrated humidity, using a calibrated background, using a calibrated spectrometer, and / or the like.

[0156] The calibrated values ​​for the calibration conditions (such as a calibrated distance, a calibrated temperature, a calibrated pressure, a calibrated humidity, a calibrated background, etc.) may include values ​​at which the model was trained and / or validated. For example, the control device 210 may receive measurement data from the spectrometer 220 that identifies values ​​of the measurement conditions (such as temperature, a distance between an unknown sample and an optical component of the spectrometer 220, etc.), and the control device 210 may validate that the model was trained using a training set and / or validation set associated with calibrated values ​​for the calibration conditions that are within a threshold amount of values.

[0157] Additionally or alternatively, the control device 210 may perform a sanity check using a single-class SVM (SC-SVM) classifier technique to determine whether the unknown spectrum is associated with a correctly performed measurement. For example, the control device 210 may aggregate multiple classes in the classification model to form an aggregated classification model having a single class, and use an SVM classifier with a decision value to determine whether the unknown sample is an outlier sample. In this case, when the unknown sample is an outlier sample, the control device 210 may determine that the set of spectral measurements was not accurately performed, and may cause the set of spectral measurements to be performed again, and may receive another set of results identifying another set of spectral measurements (block 620 - No). In this manner, the control device 210 is able to identify unknown spectra that differ from the classification model by a threshold amount without having to train the classification model using samples similar to the unknown sample (e.g., also differing by a threshold amount from the training set samples of the material of interest). Additionally, the control device 210 reduces the amount of samples to be collected for generating a classification model, thereby reducing cost, time, and computing resource utilization (e.g., processing resources and memory resources) relative to acquiring, storing, and processing additional samples of interfering material that differs from the material of interest by a threshold amount.

[0158] Furthermore, the control device 210 reduces the likelihood of inaccurate results (e.g., inaccurate quantification, inaccurate determination, etc.) from the spectroscopic method, relative to performing the method without certainty as to whether the measurement conditions match the calibration conditions. Furthermore, by determining that the measurement of the unknown sample was correctly performed before attempting to classify the unknown sample, the control device 210 reduces the utilization of computing resources, relative to attempting to perform the spectroscopic method that fails due to an incorrect measurement and then attempting another spectroscopic method.

[0159] like Figure 6As further shown in FIG6 , based on determining that the set of spectral measurements was accurately performed (block 620 — yes), process 600 may include determining whether the unknown sample is included in a mismatch category based on the results of the set of spectral measurements (block 630). For example, control device 210 may attempt (e.g., using processor 320, memory 330, storage component 340, etc.) to determine whether the unknown sample is classified as a mismatch category (e.g., a material of no interest or an interfering material). In some embodiments, control device 210 may classify the unknown sample to determine whether the unknown sample is included in the mismatch category. For example, control device 210 may use an SVM-rbf kernel function or an SVM-linear kernel function on a model to determine a decision value for classifying the unknown sample as a mismatch category. Based on the decision value satisfying a threshold decision value, control device 210 may determine that the unknown sample belongs to the mismatch category (e.g., the unknown sample is determined to be an interfering material, the spectrum is determined to be associated with a baseline spectral measurement, such as a measurement performed using an incorrect measurement distance, a measurement performed using an incorrect measurement background, a measurement performed using incorrect measurement lighting, a measurement performed in the absence of a sample, etc.). In this way, the control device 210 determines that the classification model for spectroscopy is not calibrated for the spectrum of the specific unknown sample and avoids false positive identification of the specific unknown sample. Alternatively, the control device 210 may determine that the unknown sample does not belong to the mismatch category.

[0160] like Figure 6 As further shown in FIG, based on determining that the unknown sample is included in the no-match category (block 630—yes), process 600 may include providing an output indicating that the unknown sample is included in the no-match category (block 640). For example, the control device 210 may provide information indicating that the unknown sample is included in the no-match category, for example, via a user interface (e.g., using the processor 320, the memory 330, the storage component 340, the communication interface 370, etc.). In some embodiments, the control device 210 may provide information associated with identifying the unknown sample. For example, based on attempting to quantify the amount of a particular chemical in a particular plant and determining that the unknown sample is not the particular plant (but rather another plant, e.g., based on human error), the control device 210 may provide information identifying the other plant. In some embodiments, the control device 210 may obtain another classification model and may use the other classification model to identify the unknown sample based on assigning the unknown spectrum to the no-match category of the classification model.

[0161] In this manner, the control device 210 reduces the likelihood of providing incorrect information based on a false positive identification of an unknown sample and enables the technician to perform error correction by providing information to assist in determining that the unknown sample is another plant rather than a specific plant.

[0162] like Figure 6 As further shown in FIG, based on determining that the unknown sample is not included in the unmatched category (block 630—No), process 600 may include performing one or more spectral measurements based on the results of the set of spectral measurements (block 650). For example, control device 210 may perform one or more spectral measurements based on the results of the set of spectral measurements (e.g., using processor 320, memory 330, storage component 340, etc.). In some embodiments, control device 210 may assign the unknown sample to a particular category from a set of categories of a global classification model to perform the first determination. For example, control device 210 may determine, based on the global classification model, that a particular spectrum associated with the particular sample corresponds to a local category of material (e.g., a cellulosic material, a lactose material, a caffeine material, etc.).

[0163] In some embodiments, the control device 210 may assign a particular sample based on a confidence metric. For example, the control device 210 may determine, based on the global classification model, the probability that a particular spectrum is associated with each category of the global classification model. In this case, the control device 210 may assign the unknown sample to a particular local category based on a specific probability for the particular local category exceeding other probabilities associated with other non-local categories. In this manner, the control device 210 determines the type of material associated with the sample, thereby identifying the sample. In some embodiments, the control device 210 may determine that the unknown sample does not meet a threshold associated with any category and does not meet a threshold associated with an unmatched category. In this case, the control device 210 may provide an output indicating that the unknown sample is not included in any category and cannot be assigned to the unmatched category with a confidence level corresponding to the threshold associated with the unmatched category.

[0164] In some embodiments, to perform in-situ local modeling, for example, for a classification model with a number of categories greater than a threshold, the control device 210 may generate a local classification model based on the first measurement. The local classification model may be an in-situ classification model generated using an SVM measurement technique (e.g., a kernel function such as SVM-rbf, SVM-linear, etc.; a probability-based SVM; a decision-based SVM; and / or similar techniques) based on a confidence metric associated with the first measurement. In some embodiments, the control device 210 may generate multiple local classification models.

[0165] In some embodiments, the control device 210 can generate a local quantification model based on performing a first measurement using the global classification model. For example, when the control device 210 is used to determine the concentration of a substance in an unknown sample, and multiple unknown samples are associated with different quantification models for determining the concentration of the substance, the control device 210 can use the first measurement to select a subset of categories as local categories for the unknown samples, and can select a quantification model for the unknown samples based on the results of the first measurement. In this way, the control device 210 utilizes hierarchical measurements and quantification models to improve spectral classification.

[0166] In some embodiments, the control device 210 may perform a second determination based on the results and the local classification model. For example, the control device 210 may classify the unknown sample as one of the materials of interest of the global classification model based on the local classification model and the specific spectrum. In some embodiments, the control device 210 may determine a set of confidence metrics associated with the specific spectrum and the local classification model. For example, the control device 210 may determine the probability that the specific spectrum is associated with each category of the local classification model and may assign the specific spectrum (e.g., the unknown sample associated with the specific spectrum) to a category with a higher probability than other categories of the local classification model. In this manner, the control device 210 identifies the unknown sample. In some embodiments, the control device 210 may determine a mismatch category for the local classification model and may assign the specific spectrum to the mismatch category for the local classification model. In some embodiments, the control device 210 may determine that the unknown sample fails to meet a threshold confidence metric for a category of the classification model and may determine that the classification for the unknown sample has failed. In this manner, based on using the threshold confidence metric, the control device 210 reduces the likelihood of false positive identification of the unknown sample.

[0167] In some embodiments, the control device 210 may perform quantification after performing the first determination (and / or after performing the second determination). For example, the control device 210 may select a local quantification model based on performing one or more determinations, and may perform quantification associated with a specific sample based on selecting the local quantification model. As an example, when performing raw material identification to determine the concentration of a specific chemical in a plant material, where the plant material is associated with multiple quantification models (e.g., related to whether the plant is grown indoors or outdoors, in winter or in summer, etc.), the control device 210 may perform a set of determinations to identify the specific quantification model. In this case, the control device 210 may determine that the plant is grown indoors in winter based on performing a set of determinations, and may select a quantification model associated with plants grown indoors in winter for use in determining the concentration of the specific chemical.

[0168] like Figure 6As further shown in FIG6 , upon a classification failure when performing one or more spectral classifications (block 650-A), process 600 may include providing an output indicating the classification failure and optionally updating the classification model's classification (block 660). For example, control device 210 may (e.g., using processor 320, memory 330, storage component 340, communication interface 370, etc.) provide information indicating the classification failure. For example, upon determining that a confidence level associated with the classification does not meet a threshold confidence level, control device 210 may provide an output indicating the classification failure, thereby reducing the likelihood of a false positive determination. Additionally or alternatively, upon determining that the confidence level does not meet the threshold, control device 210 may selectively update the classification model's classification for performing the classification. For example, control device 210 may obtain additional information identifying the sample (e.g., from an operator, a database, etc.) and may determine that the sample belongs to a labeled class. In this case, control device 210 may update the labeled class to enable improved subsequent spectral classifications. Additionally or alternatively, control device 210 may obtain information indicating that the sample does not belong to the labeled class. In this case, the control device 210 may update the mismatch category to enable improved subsequent mismatch classification.In this way, the control device 210 enables iterative model enhancement for spectral classification.

[0169] like Figure 6 As further shown in FIG, based on the classification success when performing one or more spectral classifications (block 650-B), process 600 may include providing information identifying a classification associated with the unknown sample (block 670). For example, control device 210 may (e.g., using processor 320, memory 330, storage component 340, communication interface 370, etc.) provide information identifying a classification associated with the unknown sample. In some embodiments, control device 210 may provide information identifying a particular classification for the unknown sample. For example, control device 210 may provide information indicating that a particular spectrum associated with the unknown sample is determined to be associated with a particular classification, thereby identifying the unknown sample.

[0170] In some embodiments, the control device 210 can provide information indicating a confidence measure associated with assigning an unknown sample to a particular class. For example, the control device 210 can provide information identifying a probability of an unknown sample being associated with a particular class, etc. In this manner, the control device 210 provides information indicating the likelihood that a particular spectrum is accurately assigned to a particular class.

[0171] In some embodiments, the control device 210 can provide quantification based on performing a set of classifications. For example, based on identifying a local quantification model associated with the class of the unknown sample, the control device 210 can provide information identifying the concentration of a substance in the unknown sample. In some embodiments, the control device 210 can update a classification model based on performing a set of classifications. For example, based on determining the classification of the unknown sample as a material of interest, an interfering material, and / or a similar material, the control device 210 can generate a new classification model that includes the unknown sample as a training set of samples.

[0172] Although Figure 6 Example blocks of process 600 are shown, but in some implementations, Figure 6 Process 600 may include additional blocks, fewer blocks, different blocks, or differently arranged blocks compared to those depicted in . Additionally or alternatively, two or more blocks of process 600 may be executed in parallel.

[0173] Figure 7A and Figure 7B is with Figure 6 FIG. 7 is a diagram of an example implementation 700 associated with predicting success rates associated with the example process 600 shown in FIG. Figure 7A and Figure 7B Example results for stock identification using a technique based on hierarchical support vector machines (Hierarchical-SVM Linear) are shown.

[0174] like Figure 7A As shown by reference numeral 705, the control device 210 may cause the spectrometer 220 to perform a set of spectral measurements. For example, the control device 210 may provide instructions to cause the spectrometer 220 to obtain a spectrum of an unknown sample to determine the classification of the unknown sample as a particular material of interest from a set of materials of interest that the classification model is trained to identify. As shown by reference numerals 710 and 715, the spectrometer 220 may receive the unknown sample and may perform a set of spectral measurements on the unknown sample. As shown by reference numeral 720, the control device 210 may receive a spectrum of the unknown sample based on the set of spectral measurements performed by the spectrometer 220 on the unknown sample.

[0175] like Figure 7B As shown, the control device 210 can perform spectral classification using a classification model 725. The classification model 725 includes a set of categories 730 identified for a set of spectra in the training set. For example, the classification model 725 includes categories 730-1 through 730-6 for materials of potential interest and a mismatch category 730-7 for interfering materials (e.g., similar materials; similar spectra; incorrectly obtained spectra, such as incorrect lighting spectra, incorrect distance spectra, incorrect background spectra, etc.; and / or the like).

[0176] like Figure 7B As further illustrated by reference numerals 735 and 740 in FIG, the spectrum of the unknown sample is assigned to a mismatch class, and the unknown sample is identified as an interfering material (e.g., a member of the mismatch class). For example, the control device 210 can use an in-situ local modeling technique to generate a local model based on a global model (e.g., classification model 725), and can determine whether the unknown sample is an interfering material based on the local model. In some embodiments, the control device 210 can perform an in-situ thresholding technique to determine whether the unknown sample is an interfering material. For example, the client device 750 can self-validate or cross-validate a decision value associated with the first most likely class of the unknown sample and / or the runner-up class (e.g., the second most likely class) of the sample, and can use the decision value to set upper and lower limits for the prediction threshold. In some embodiments, the client device 750 can utilize multiple local modeling strategies. For example, the client device 750 can utilize a first modeling technique to determine a winner class and a second modeling technique to determine a confidence measure. In some embodiments, the client device 750 can utilize a single-class support vector machine (SC-SVM) technique to determine whether the unknown sample is an interfering material. As shown by reference numeral 745 , the control device 210 provides an output to the client device 750 indicating that the unknown sample is an interfering material, rather than providing a false positive identification of the unknown sample as a material of interest at a particular concentration within the material of interest.

[0177] As indicated above, Figure 7A and Figure 7B It is provided as an example only. Other examples are possible and may differ from the Figure 7A and Figure 7B An example of description.

[0178] In this manner, the control device 210 reduces the likelihood of providing inaccurate results for the spectroscopy method based on avoiding false positive identification of unknown samples as a particular material of interest that the classification model is trained to identify.

[0179] The foregoing disclosure provides illustration and description, but is not intended to be exhaustive or to limit the embodiments to the precise form disclosed. Modifications and variations are possible in light of the above disclosure or may be acquired from practice of the embodiments.

[0180] Some embodiments are described herein in conjunction with threshold values. As used herein, satisfying a threshold value may refer to a value being greater than a threshold value, more than a threshold value, higher than a threshold value, greater than or equal to a threshold value, less than a threshold value, less than a threshold value, lower than a threshold value, less than or equal to a threshold value, equal to a threshold value, etc.

[0181] It will be apparent that the systems and / or methods described herein can be implemented in various forms of hardware, firmware, or a combination of firmware and software. The actual dedicated control hardware or software code used to implement these systems and / or methods is not a limitation of the embodiments. Therefore, the operation and behavior of the systems and / or methods are described herein without reference to specific software code, with the understanding that software and hardware can be designed to implement the systems and / or methods based on the description herein.

[0182] Although particular combinations of features are recited in the claims and / or disclosed in the specification, these combinations are not intended to limit the disclosure of possible implementations. In fact, many of these features can be combined in ways not specifically recited in the claims and / or disclosed in the specification. Although each of the attached dependent claims may be directly dependent on only one claim, the disclosure of possible implementations includes each dependent claim in combination with every other claim in the claim set.

[0183] Any element, action or instruction used in this article should not be interpreted as key or necessary, unless clearly described in this way. In addition, as used in this article, the article "a" and "an" are intended to include one or more projects, and can be used interchangeably with "one or more". In addition, as used in this article, the term "set" is intended to include one or more projects (for example, related projects, unrelated projects, the combination of related projects and unrelated projects, etc.), and can be used interchangeably with "one or more". In the case of meaning only one project, the term "one" or similar language is used. In addition, as used in this article, the term "has", "have", "having", and / or similar words are intended to be open terms. In addition, unless otherwise explicitly stated, the phrase "based on" is intended to represent "at least partially based on".

Claims

1. A method for reduced false positive identification for spectral classification, comprising: determining, by the device, based on information identifying a particular result of a spectral measurement of the unknown sample, that the unknown sample is not included in a mismatch category associated with at least one of at least one material not of interest or a baseline spectral measurement, wherein the baseline spectral measurement corresponds to at least one of: a measurement performed using an incorrect measurement distance, a measurement performed using an incorrect measurement background, a measurement performed using incorrect measurement illumination, and a measurement performed without a sample present; performing, by the device, one or more spectroscopic measurements based on determining that the unknown sample is not included in the no-match category; determining, by the device and based on performing the one or more spectroscopic measurements, a classification failure or a classification success for the unknown sample; as well as One or more actions are performed by the device based on determining a classification failure or a classification success for the unknown sample.

2. The method according to claim 1, further comprising: Before determining that the unknown sample is not included in the unmatched class, it is determined that the spectral measurement of the unknown sample is accurately performed.

3. The method according to claim 2, wherein: Determining that the spectral measurement of the unknown sample is performed accurately includes: Determine that the spectral measurements of the unknown sample are performed at the following calibrated distances: the calibrated distance between the spectrometer's optical components and the unknown sample, or A calibrated distance between the optical components of the spectrometer and the background of the unknown sample.

4. The method according to claim 1, further comprising: determining a decision value for classifying the unknown sample into the unmatched category; as well as determining that the decision value does not meet a threshold decision value, Wherein, determining that the unknown sample is not included in the unmatched category includes: Based on determining that the decision value does not satisfy the threshold decision value, the unknown sample is determined not to be included in the unmatched class.

5. The method according to claim 4, wherein Determining the decision value includes: The decision value is determined using a support vector machine-radial basis function (SVM-rbf) kernel function or an SVM-linear kernel function.

6. The method according to claim 1, wherein Performing the one or more spectroscopic measurements comprises: The unknown sample is assigned to a particular class from a set of classes of a global classification model to perform the one or more spectrometric determinations.

7. The method according to claim 1, wherein Performing the one or more spectroscopic measurements comprises: performing a first determination of the one or more spectrometric determinations using the global classification model; identifying a local classification model for the unknown sample based on performing the first determination; and Based on the results of the first determination and the local classification model, a second determination of the one or more spectrometric determinations is performed.

8. The method according to claim 1, wherein Performing the one or more actions includes: Based on determining that the classification fails, the marked category or the mismatch category is updated.

9. The method according to claim 1, wherein Performing the one or more actions includes: Based on determining a classification success for the unknown sample, information is provided identifying a probability that the unknown sample is associated with a particular class.

10. A device for reduced false positive identification of spectral classification, comprising: one or more memories; as well as one or more processors communicatively coupled to the one or more memories, the one or more processors configured to: determining, based on information identifying a particular result of a spectral measurement of the unknown sample, that the unknown sample is not included in a mismatch category associated with at least one of at least one material not of interest or a baseline spectral measurement, wherein the baseline spectral measurement corresponds to at least one of: a measurement performed using an incorrect measurement distance, a measurement performed using an incorrect measurement background, a measurement performed using incorrect measurement illumination, and a measurement performed without a sample present; performing one or more spectroscopic measurements based on determining that the unknown sample is not included in the no-match category; as well as Based on performing the one or more spectroscopic measurements, one or more actions are performed.

11. The apparatus according to claim 10, wherein The one or more processors are further configured to: Before determining that the unknown sample is not included in the unmatched class, it is determined that the spectral measurement of the unknown sample is accurately performed.

12. The apparatus according to claim 11, wherein The one or more processors, upon determining that the spectral measurement of the unknown sample is accurately performed, are configured to: Determine that the spectral measurements of the unknown sample are performed at the following calibrated distances: the calibrated distance between the spectrometer's optical components and the unknown sample, or A calibrated distance between the optical components of the spectrometer and the background of the unknown sample.

13. The apparatus according to claim 10, wherein The one or more processors are further configured to: determining a decision value for classifying the unknown sample into the unmatched category; and determining that the decision value does not meet a threshold decision value, Wherein, when determining that the unknown sample is not included in the unmatched category, the one or more processors are configured to: Based on determining that the decision value does not satisfy the threshold decision value, the unknown sample is determined not to be included in the unmatched class.

14. The apparatus according to claim 13, wherein The one or more processors, when determining the decision value, are configured to: The decision value is determined using a support vector machine-radial basis function (SVM-rbf) kernel function or an SVM-linear kernel function.

15. The apparatus according to claim 10, wherein The one or more processors, when performing the one or more spectrometric measurements, are configured to: The unknown sample is assigned to a particular class from a set of classes of a global classification model to perform the one or more spectrometric determinations.

16. A non-transitory computer-readable medium storing a set of instructions, the set of instructions comprising: One or more instructions that, when executed by one or more processors, cause the one or more processors to: determining that the unknown sample is not included in the no-match category based on the information identifying the specific result of the spectral measurement of the unknown sample; performing one or more spectroscopic measurements based on determining that the unknown sample is not included in the no-match category; determining a classification failure or a classification success for the unknown sample based on performing the one or more spectroscopic measurements; as well as Based on determining a classification failure or a classification success for the unknown sample, one or more actions are performed.

17. The non-transitory computer readable medium of claim 16, wherein: The one or more instructions further cause the one or more processors to: Before determining that the unknown sample is not included in the unmatched class, it is determined that the spectral measurement of the unknown sample is accurately performed.

18. The non-transitory computer readable medium of claim 16, wherein: The mismatch category is associated with at least one of at least one material of no interest or a baseline spectral measurement, and the baseline spectral measurement corresponds to at least one of: a measurement performed using an incorrect measurement distance, a measurement performed using an incorrect measurement background, a measurement performed using incorrect measurement illumination, and a measurement performed in the absence of a sample.

19. The non-transitory computer readable medium of claim 16, wherein: The one or more instructions further cause the one or more processors to: determining a decision value for classifying the unknown sample into the unmatched category; and determining that the decision value does not meet a threshold decision value, wherein the one or more instructions causing the one or more processors to determine that the unknown sample is not included in the no-match category cause the one or more processors to: Based on determining that the decision value does not satisfy the threshold decision value, the unknown sample is determined not to be included in the unmatched class.

20. The non-transitory computer readable medium of claim 19, wherein: The one or more instructions causing the one or more processors to determine the decision value cause the one or more processors to: The decision value is determined using a support vector machine-radial basis function (SVM-rbf) kernel function or an SVM-linear kernel function.

Images