HK1232956B - Identification using spectroscopy - Google Patents

Description

Identification using spectroscopy

background

Raw material identification can be used for quality control of pharmaceutical products. For example, raw material identification can be performed on a medicinal compound to determine whether the composition of the medicinal compound corresponds to the packaging label associated with the medicinal compound. Spectroscopy can facilitate non-destructive raw material identification by taking advantage of reduced preparation and data acquisition time compared to other chemical techniques.

Overview

According to some possible implementations, the device may include one or more processors. The one or more processors may receive information identifying the results of spectral measurements of unknown samples. The one or more processors may perform a first classification of the unknown sample based on the results of the spectral measurements and a global classification model. The global classification model may utilize support vector machine (SVM) classifier technology. The global classification model may include a global group of categories. The one or more processors may generate a local classification model based on the first classification. The local classification model may utilize SVM classifier technology. The local classification model may include a subset of categories of the global group of categories. The one or more processors may perform a second classification of the unknown sample based on the results of the spectral measurements and the local classification model. The one or more processors may provide information identifying categories associated with the unknown sample in a subset of categories based on performing the second classification.

According to some possible implementations, a computer-readable medium may store instructions that, when executed by one or more processors, may cause the one or more processors to receive information identifying the results of a set of spectral measurements of an unknown group. The unknown group may include a set of unknown samples. The one or more instructions, when executed by one or more processors, may cause the one or more processors to perform a first classification of the set of unknown samples based on the results of the set of spectral measurements and a global classification model. The global classification model may utilize support vector machine (SVM) linear classifier technology. The one or more instructions, when executed by one or more processors, may cause the one or more processors to generate a set of local classification models for the set of unknown samples based on the first classification. The set of local classification models may utilize SVM linear classifier technology. The one or more instructions, when executed by one or more processors, may cause the one or more processors to perform a second classification of the set of unknown samples based on the results of the set of spectral measurements and the set of local classification models. The one or more instructions, when executed by one or more processors, may cause the one or more processors to provide information identifying the classification of the set of unknown samples based on performing the second classification.

According to some possible implementations, the method may include receiving, by a device, information identifying a result of a spectral measurement of an unknown sample performed by a first spectrometer. The method may include performing, by the device, a first classification of the unknown sample based on the result of the spectral measurement and a global classification model. The global classification model may be generated by utilizing support vector machine (SVM) classifier technology and a set of spectral measurements performed by a second spectrometer. The method may include generating, by the device, a local classification model based on the first classification. The local classification model may utilize SVM classifier technology. The local classification model may include a subset of categories from a set of categories of the global classification model. The method may include performing, by the device, a second classification of the unknown sample based on the result of the spectral measurement and the local classification model. The method may include providing, by the device, information identifying a category associated with the unknown sample from a subset of categories based on performing the second classification.

(1) This application relates to a device comprising:

One or more processors configured to:

receiving information identifying results of a spectral measurement of an unknown sample;

Perform a first classification of unknown samples based on the results of spectral measurements and a global classification model,

Generate a local classification model based on the first classification,

performing a second classification of the unknown sample based on the results of the spectral measurement and the local classification model; and

Information identifying a class associated with the unknown sample is provided based on performing the second classification.

(2) The device as described in (1), wherein the one or more processors are further configured to:

determining a corresponding set of probabilities associated with a set of categories for a global classification model,

A particular probability in a set of corresponding probabilities indicates the likelihood that the unknown sample is associated with a particular class in a set of classes,

selecting a subset of a set of classes based on a set of corresponding probabilities; and

One or more of the processors are used to generate a local classification model:

Generate a local classification model based on a subset of a set of categories.

(3) The device as described in (1), wherein the one or more processors are further configured to:

performing an automatic scaling preprocessing process; and

At least one of the first classification or the second classification is performed based on performing an automatic scaling pre-processing process.

(4) The device as described in (1), wherein the one or more processors are further configured to:

receiving a global classification model from a control device associated with the first spectrometer,

The global classification model is generated by the control device using one or more spectral measurements performed by the first spectrometer;

so that the spectral measurement will be performed by a second spectrometer,

The second spectrometer is different from the first spectrometer; and

The one or more processors, when performing a first classification of an unknown sample, are configured to:

A first classification of the unknown sample is performed based on a global classification model generated using one or more spectral measurements performed by the first spectrometer and based on results of spectral measurements performed by the second spectrometer.

(5) The apparatus of (1), wherein the set of categories of the global classification model corresponds to a set of compounds, and the category is included in the set of categories; and

One or more of these processors, when providing the identified categories of information, is used to:

Provides information identifying compounds within a group of compounds that correspond to a class.

(6) The apparatus of (1), wherein the one or more processors, when performing the second classification, are configured to:

Determine the spectra associated with the unknown sample and the class associated with it,

The spectrum is identified as a result of performing a spectral measurement; and

One or more of these processors, when providing the identified categories of information, is used to:

Information identifying the class is provided based on determining that the spectrum associated with the unknown sample is associated with the class.

(7) The apparatus of (1), wherein a support vector machine (SVM) classifier technique is used to generate at least one of the global classification model or the local classification model; and

Wherein the SVM classifier technique is associated with at least one of the following:

Kernel functions of the radial basis function type,

Kernel function of linear function type,

S-type kernel function,

a kernel function of polynomial type, or

Exponential kernel function.

(8) The apparatus of (1), wherein the one or more processors, when performing the second classification, are configured to:

Assign unknown samples to categories based on at least one of the following:

probability value, or

Decision value.

(9) The present application provides a computer-readable medium storing instructions, the instructions comprising:

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

receiving information identifying results of a set of spectral measurements of an unknown group,

The unknown group includes a plurality of unknown samples;

performing a first classification of a plurality of unknown samples based on the results of a set of spectral measurements and a global classification model,

The global classification model uses support vector machine (SVM) linear classifier technology.

generating a set of local classification models for a plurality of unknown samples based on the first classification,

The set of local classification models utilizes SVM linear classifier technology;

performing a second classification of the plurality of unknown samples based on the results of the set of spectral measurements and the set of local classification models; and

Information identifying classifications of the plurality of unknown samples is provided based on performing the second classification.

(10) The computer-readable medium of (9), wherein the global classification model is generated based on one or more spectral measurements performed by the first spectrometer; and

Wherein the one or more instructions causing the one or more processors to receive information identifying results of a set of spectral measurements cause the one or more processors to:

receiving information identifying results of a set of spectral measurements from a second spectrometer,

The second spectrometer is different from the first spectrometer; and

wherein causing the one or more processors to execute the one or more instructions of the first classification causes the one or more processors to:

The first classification is performed using results of a set of spectral measurements received from the second spectrometer and a global classification model generated based on one or more spectral measurements performed by the first spectrometer.

(11) The computer-readable medium of (9), wherein the one or more instructions that cause the one or more processors to receive information identifying results of a set of spectral measurements cause the one or more processors to:

receiving a plurality of spectra corresponding to a plurality of unknown samples; and

wherein causing the one or more processors to execute the one or more instructions of the first classification causes the one or more processors to:

Assign multiple spectra to one or more classes of a global classification model,

One or more classes of the global classification model correspond to one or more compounds.

12. The computer-readable medium of claim 11 , wherein the one or more instructions, when executed by the one or more processors, further cause the one or more processors to:

determining one or more confidence measures for the one or more classes and a particular spectrum of the plurality of spectra,

A confidence metric of the one or more confidence metrics indicates a likelihood that a particular spectrum is associated with a characteristic class of the one or more classes corresponding to the confidence metric;

assigning a particular spectrum to a particular class based on one or more confidence metrics; and

wherein the one or more instructions causing the one or more processors to generate a set of local classification models causes the one or more processors to:

selecting a subset of the one or more categories based on the one or more confidence metrics; and

A specific local classification model is generated based on a subset of one or more categories to form a set of local classification models.

(13) The computer-readable medium of (9), wherein the one or more instructions that cause the one or more processors to receive information identifying results of a set of spectral measurements cause the one or more processors to:

receiving a plurality of spectra corresponding to a plurality of unknown samples; and

wherein causing the one or more processors to execute the one or more instructions of the second classification causes the one or more processors to:

Assign multiple spectra to one or more classes of a set of local classification models,

One or more classes of a set of local classification models correspond to one or more compounds; and

wherein the one or more instructions causing the one or more processors to provide for identifying a classification of the plurality of unknown samples further:

Information indicating one of the one or more categories to which a spectrum in the plurality of spectra associated with one of the unknown samples is assigned is provided based on assigning the plurality of spectra to the one or more categories.

(14) The computer-readable medium of (9), wherein the one or more instructions that cause the one or more processors to execute the second classification cause the one or more processors to:

determining a set of decision values associated with a particular local classification model from a set of local classification models and a particular unknown sample from a plurality of unknown samples,

The decision value corresponds to one of a set of classes of a particular local classification model; and

Assigns a particular unknown sample to one of a set of classes based on a set of decision values.

(15) The present application relates to a method comprising:

receiving, by the device, information identifying results of a spectral measurement of the unknown sample performed by the first spectrometer;

A first classification of the unknown sample is performed by the device based on the results of the spectral measurements and the global classification model,

A global classification model is generated using a support vector machine (SVM) classifier technique and a set of spectral measurements performed by a second spectrometer;

generating, by the device, a local classification model based on the first classification,

The local classification model uses SVM classifier technology.

The local classification model includes a subset of the categories from the set of categories of the global classification model;

performing, by the device, a second classification of the unknown sample based on the results of the spectral measurement and the local classification model; and

Information identifying a class in the subset of classes associated with the unknown sample is provided by the device based on performing the second classification.

(16) The method of (15), wherein the kernel function associated with the SVM classifier technique includes at least one of the following:

Kernel functions of the radial basis function type,

Kernel function of linear function type,

S-type kernel function,

a kernel function of polynomial type, or

Exponential kernel function.

(17) The method of (15), wherein performing the second classification comprises:

Assign unknown samples to a subset of classes based on at least one of the following:

The probability value associated with the class, or

The decision value associated with the category.

(18) The method of (15), wherein the first spectrometer is different from the second spectrometer.

(19) The method as described in (15), further comprising:

receiving a global classification model from a control device associated with a second spectrometer;

storing the global classification model via a data structure; and

The first category includes:

deriving a global classification model from the data structure; and

The first classification is performed using the global classification model.

(20) The method as described in (15), further comprising:

providing information identifying a confidence measure associated with the second classification,

The confidence metric represents a measure of the confidence with which an unknown sample is assigned to a class.

BRIEF DESCRIPTION OF THE DRAWINGS

1A and 1B are diagrams of overviews of example implementations described herein;

FIG2 is an illustration of an example environment in which the systems and/or methods described herein may be implemented;

FIG3 is a diagram of example components of one or more devices of FIG2;

4 is a flow chart of an example process for generating a global classification model for raw material identification based on a support vector machine classifier;

FIG5 is a diagram of an example implementation related to the example process shown in FIG4;

FIG6 is a flow chart of an example process for performing raw material identification using a multi-level classification technique; and

7A and 7B are diagrams of example implementations of predicting success rates associated with the example process shown in FIG. 6 .

Detailed description

The following detailed description of example implementations refers to the accompanying drawings, in which the same reference numerals in different drawings may identify the same or similar elements.

Raw material identification (RMID) is a technique for identifying the components (e.g., ingredients) of a particular sample for identification, verification, and the like. For example, RMID can be used to verify that the components in a pharmaceutical compound correspond to a set of components identified on a label. A spectrometer can be used to perform spectroscopy on a sample (e.g., a pharmaceutical compound) to determine the components of the sample. The spectrometer can determine a set of measurements of the sample and can provide the set of measurements for classification. Chemometric classification techniques (e.g., classifiers) can facilitate determination of the components of a sample based on the set of measurements of the sample. However, relative to other techniques, some chemometric classification techniques may be associated with poor transferability, insufficient granularity for performing large-scale classification, and the like. The implementation described herein can utilize a hierarchical support vector machine classifier to facilitate RMID. In this way, relative to other RMID techniques, the control device of the spectrometer contributes to improved classification accuracy.

Figures 1A and 1B are diagrams of overviews of an example implementation 100 described herein. As shown in Figure 1A, the example implementation 100 may include a first control device and a first spectrometer. The first control device may cause the first spectrometer to perform a set of spectral measurements on a training set (e.g., a set of known samples used to train a classification model). The training set may be selected to include a threshold number of samples for each class of the classification model. The classes of the classification model may relate to a group of similar compounds that share one or more common features, such as (in a pharmaceutical context) lactose compounds, fructose compounds, acetaminophen compounds, ibuprofen compounds, aspirin compounds, etc.

As further shown in Figure 1A, the first spectrometer may perform a set of spectral measurements on the training set based on instructions received from the first control device. For example, the first spectrometer may determine a spectrum for each sample in the training set. The first spectrometer may provide the set of spectral measurements to the first control device. The first control device may use a specific classification technique and generate a global classification model based on the set of spectral measurements. For example, the first control device may use support vector machine (SVM) technology (e.g., a machine learning technology for information classification) to generate the global classification model. The global classification model may include information related to assigning specific spectra to specific categories, and may include information related to identifying the type of compound associated with the specific category. In this way, the control device can provide information to identify the type of compound of the unknown sample based on assigning the spectrum of the unknown sample to a specific category. The global classification model may be stored via a data structure, provided to one or more other control devices, and so on.

As shown in FIG1B , a second control device may receive a global classification model (e.g., from a first control device) and may store the global classification model via a data structure. The second control device may cause a second spectrometer to perform a set of spectral measurements on an unknown group (e.g., a group of unknown samples for which RMID is to be performed). The second spectrometer may perform the set of spectral measurements based on instructions received from the second control device. For example, the second spectrometer may determine a spectrum for each sample in the unknown group. The second spectrometer may provide the set of spectral measurements to the second control device. The second control device may perform RMID on the unknown group using a multi-level classification technique based on the global classification model.

With respect to FIG. 1B , a second control device may use a global classification model to perform a first classification of a specific sample of an unknown group. The second control device may determine a set of confidence metrics associated with the specific sample and the global classification model. The confidence metrics may relate to the confidence associated with assigning the specific sample to a specific category. For example, the second control device may determine a confidence metric associated with the specific sample and each category of the global classification model. The second 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 involve an in-situ classification model generated using support vector machines (SVMs) techniques and a subset of categories. The second control device may perform a second classification based on the local classification model to assign the specific sample to a specific category. In this manner, the second control device performs RMID on the specific sample of the unknown group with improved accuracy relative to other classification models and/or single-pole classification techniques. The second control device may perform the first classification and the second classification on each sample of the unknown group to identify each sample of the unknown group. In another example, the first control device may use the global classification model and the local classification model to classify the specific sample based on spectroscopy performed by the first spectrometer.

FIG2 is a diagram of an example environment 200 in which the systems and/or methods described herein may be implemented. As shown in FIG2 , 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.

The control device 210 may include one or more devices capable of storing, processing, and/or routing information related to RMID. For example, the control device 210 may include a server, a computer, a wearable device, a cloud computing device, etc., which generates a model based on a classifier and a set of measurements from a training set and uses the model to perform RMID based on a set of measurements from an unknown set. In some implementations, the control device 210 may be associated with a specific spectrometer 220. In some implementations, the control device 210 may be associated with multiple spectrometers 220. In some implementations, the control device 210 may receive information from another device in the environment 200 (e.g., the spectrometer 220) and/or send information to another device in the environment 200 (e.g., the spectrometer 220).

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 a near-infrared (NIR) spectrometer, mid-infrared spectroscopy (mid-IR), Raman spectroscopy, etc.). In some implementations, spectrometer 220 may be incorporated into a wearable device, such as a wearable spectrometer. In some implementations, spectrometer 220 may receive information from another device in environment 200 (such as control device 210) and/or send information to another device in environment 200 (such as control device 210).

The network 230 may include one or more wired 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 combinations of these and other types of networks.

The number and arrangement of the devices and networks shown in FIG2 are provided as examples. In practice, 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 compared to those shown in FIG2 . In addition, two or more devices shown in FIG2 may be implemented within a single device, or the single device shown in FIG2 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 (e.g., one or more devices) of environment 200 may perform one or more functions described as being performed by another group of devices of environment 200.

FIG3 is a diagram of example components of a device 300. Device 300 may correspond to control device 210 and/or spectrometer 220. In some implementations, control device 210 and/or spectrometer 220 may include one or more devices 300 and/or one or more components of device 300. As shown in FIG3 , 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.

The bus 310 may include 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 may include a processor (e.g., a central processing unit (CPU), a graphics processing unit (GPU), an accelerated processing unit (APU), etc.), a microprocessor, and/or any processing component that interprets and/or executes instructions (e.g., a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), etc.). In some implementations, the processor 320 may include one or more processors that can be programmed to perform functions. The memory 330 may include random access memory (RAM), read-only memory (ROM), and/or another type of dynamic or static storage device (e.g., flash memory, magnetic memory, optical memory, etc.) that stores information and/or instructions for use by the processor 320.

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

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

The communication interface 370 may include transceiver-like components (e.g., a transceiver, a separate receiver and transmitter, etc.) that enable the device 300 to communicate with other devices, such as via a wired connection, a wireless connection, or a combination of wired and wireless connections. The communication interface 370 may allow the device 300 to receive information from another device and/or provide information to another device. For example, the communication interface 370 may 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 Wi-Fi interface, a cellular network interface, etc.

Device 300 can perform one or more processes described herein. Device 300 can perform these processes in response to processor 320 executing software instructions stored by a computer-readable medium (such as memory 330 and/or storage component 340). Computer-readable media is defined herein as a non-transitory memory device. A memory device includes storage space within a single physical storage device or storage space spread across multiple physical storage devices.

The software instructions may be read from another computer-readable medium or from another device into the memory 330 and/or storage component 340 via the communication interface 370. When executed, the software instructions stored in the memory 330 and/or storage component 340 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 implementations described herein are not limited to any specific combination of hardware circuitry and software.

The number and arrangement of components shown in FIG3 are provided as examples. In practice, device 300 may include additional components, fewer components, different components, or differently arranged components compared to those shown in FIG3. Additionally or alternatively, one or more components of environment 300 (e.g., one or more components) may perform one or more functions described as being performed by another group of components of device 300.

FIG4 is a flow chart of an example process 400 for generating a global classification model for raw material identification based on a support vector machine classifier. In some implementations, one or more process blocks of FIG4 can be performed by the control device 210. In some implementations, one or more process blocks of FIG4 can be performed by another device or group of devices (such as the spectrometer 220) that is separate from or includes the control device 210.

As shown in Figure 4, process 400 may include causing a set of spectral measurements to be performed on a training set (block 410). For example, the control device 210 may cause the spectrometer 220 to perform a set of spectral measurements on the training set of samples to determine the spectrum for each sample of the training set. A training set may refer to a set of samples of one or more known compounds that can be used to generate a global classification model. For example, a training set may include one or more versions of a set of compounds (e.g., one or more versions manufactured by different manufacturers to control manufacturing differences). In some implementations, a training set may be selected based on the compounds of the expected group on which RMID will be performed. For example, when it is expected that RMID will be performed on a pharmaceutical compound, the training set may include a set of samples of active pharmaceutical ingredients (APIs), excipients, etc. In some implementations, a training set may be selected to include a specific number of samples for each type of compound. For example, a training set may be selected to include multiple samples of a specific compound (e.g., 5 samples, 10 samples, 15 samples, 50 samples, etc.). In this manner, the controller 210 may be provided with a threshold number of spectra associated with a particular type of compound, thereby facilitating the generation of a classification model (eg, a global classification model, a local classification model, etc.) to which an unknown sample can be accurately assigned.

In some implementations, the control device 210 can cause multiple spectrometers 220 to perform the set of spectral measurements to account for 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 that the spectral measurements will be inaccurate as a result of the physical conditions that caused the set of spectral measurements to be performed by a single spectrometer 220.

As further shown in FIG4 , process 400 may include receiving information identifying the results of the set of spectral measurements (block 420). For example, control device 210 may receive information identifying the results of the set of spectral measurements. In some implementations, control device 210 may receive information identifying a set of spectra corresponding to samples of the training set. For example, control device 210 may receive information identifying specific spectra observed when spectrometer 220 performed spectroscopy on the training set. Additionally or alternatively, control device 210 may receive other information as a result of the set of spectral measurements. For example, control device 210 may receive information related to identifying absorption of energy, emission of energy, scattering of energy, and the like.

In some implementations, the control device 210 can receive information identifying the results of the set of spectral measurements from the plurality of spectrometers 220. For example, the control device 210 can control physical conditions, such as 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.

As further shown in FIG4 , process 400 may include generating a global classification model associated with a particular classifier based on the information identifying the results of the set of spectral measurements (block 430). For example, the control device 210 may generate a global classification model associated with a SVM classifier technique based on the information identifying the results of the set of spectral measurements. In some implementations, the control device 210 may perform a set of classifications to generate the global classification model. For example, the control device 210 may assign the set of spectra identified by the results of the set of spectral measurements to a set of categories based on the use of the SVM technique.

SVM can refer to a supervised learning model that performs pattern recognition for classification. In some implementations, when using the SVM technology to generate a global classification model, the control device 210 can utilize a specific type of kernel function. For example, the control device 210 can utilize a kernel function of the radial basis function (RBF) (e.g., referred to as SVM-rbf) type, a kernel function of the linear function (e.g., when used for multi-class classification technology, referred to as SVM linear and referred to as hier-SVM linear) type, a kernel function of the S-type function type, a kernel function of the polynomial function type, a kernel function of the exponential function type, etc. In some implementations, the control device 210 can utilize a specific type of SVM, such as an SVM based on a probability value (e.g., a classification based on the probability that a sample is a member of a class in a set of classes), an SVM based on a decision value (e.g., a classification using a decision function to vote in favor of a class in a set of classes as a class of which the sample is a member), etc.

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

In some implementations, the control device 210 may generate a global classification model based on information identifying the samples of the training set. For example, the control device 210 may use information identifying the types of compounds represented by the samples of the training set to identify the categories of spectra having the types of compounds. In some implementations, when generating the global classification model, the control device 210 may train the global classification model. For example, the control device 210 may use a portion of the set of spectral measurements to train the model. Additionally or alternatively, the control device 210 may perform an evaluation of the global classification model. For example, the control device 210 may use another portion of the set of spectral measurements to validate the global classification model (e.g., for predicted intensity). In some implementations, the control device 210 may use a multi-level classification technique to validate the global classification model. For example, when one or more local classification models are utilized, the control device 210 may determine that the global classification model is accurate, as described herein with respect to FIG. 6 . In this way, the control device 210 ensures that a global classification model with a threshold accuracy is generated before providing the global classification model for utilization by other control devices 210 associated with other spectrometers 220.

In some implementations, the control device 210 may provide the global classification model to other control devices 210 associated with other spectrometers 220 after generating the global classification model. For example, a first control device 210 may generate a global classification model and may provide the global classification model to a second control device 210 for use. In this case, the second control device 210 may store the global classification model and may use the global classification model when generating one or more local classification models and classifying one or more samples of the unknown group, as described herein with respect to FIG. 6 . Additionally or alternatively, the control device 210 may store the global classification model for use by the control device 210 when generating one or more local classification models and classifying one or more samples. In this manner, the control device 210 provides the global classification model for use in the RMID of the unknown sample.

Although Figure 4 shows example blocks of process 400, in some implementations, process 400 may include additional blocks, fewer blocks, different blocks, or differently arranged blocks than those depicted in Figure 4. Additionally or alternatively, two or more blocks of process 400 may be performed in parallel.

Figure 5 is a diagram of an example implementation 500 related to the example process 400 shown in Figure 4. Figure 5 shows an example of generating a global classification model for raw material identification based on a support vector machine classifier.

As shown in Figure 5, 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 510. It is assumed that the training set 510 includes a first set of training samples (e.g., measurements thereof are used to train the global classification model) and a second set of validation samples (e.g., measurements thereof are used to verify the accuracy of the global classification model). As shown by reference numeral 515, the spectrometer 220-1 performs the set of spectral measurements on the training set 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. It is assumed that the control device 210-1 stores information identifying each sample of the training set 510.

5 , assume that control device 210-1 has selected to utilize a hier-SVM linear classifier for generating a global classification model (e.g., based on testing the hier-SVM linear classifier against one or more other classifiers). As indicated by reference numeral 525, control device 210-1 trains the global classification model using the hier-SVM linear classifier and the first set of spectra, and validates the global classification model using the hier-SVM linear classifier and the second set of spectra. Assume that control device 210-1 determines that the global classification model satisfies a validation threshold (e.g., has an accuracy that exceeds the validation threshold). As indicated by reference numeral 530, control device 210-1 provides the global classification model to control device 210-2 (e.g., for utilization when performing RMID on spectral measurements performed by spectrometer 220-2) and to control device 210-3 (e.g., for utilization when performing RMID on spectral measurements performed by spectrometer 220-3).

As indicated above, FIG5 is provided only as an example. Other examples are possible and may differ from what is described with respect to FIG5.

In this manner, the control device 210 facilitates the generation of a global classification model based on a selected classification technique (e.g., selected based on model transferability, large-scale classification accuracy, etc.) and the distribution of the global classification model for utilization by one or more other control devices 210 associated with one or more spectrometers 220. Furthermore, the control device 210 reduces the cost and time requirements relative to generating the global classification model on each control device 210 that will perform RMID.

FIG6 is a flow chart of an example process 600 for performing raw material identification using a multi-level classification technique. In some implementations, one or more process blocks of FIG6 can be performed by the control device 210. In some implementations, one or more process blocks of FIG6 can be performed by another device or group of devices, such as the spectrometer 220, that is separate from or includes the control device 210.

6 , process 600 may include receiving information identifying results of a set of spectral measurements performed on an unknown group (block 610). For example, the control device 210 may receive information identifying results of the set of spectral measurements performed by the spectrometer 220 on the unknown group. The unknown group may include a group of samples (e.g., unknown samples) for which RMID is to be performed. For example, the control device 210 may cause the spectrometer 220 to perform the set of spectral measurements on the group of samples and may receive information identifying a set of spectra corresponding to the group of samples. In some implementations, the control device 210 may receive information identifying results from multiple spectrometers 220. For example, the control device 210 may cause multiple spectrometers 220 to perform the set of spectral measurements on the unknown group (e.g., the same group of samples) and may receive information identifying a set of spectra corresponding to the samples of the unknown group. Additionally or alternatively, the control device 210 may receive information identifying results of a set of spectral measurements performed at multiple times, at multiple locations, etc., and may classify a particular sample based on the set of spectral measurements (e.g., based on averaging the set of spectral measurements or based on another technique) performed at multiple times, at multiple locations, etc. In this manner, the control device 210 may take into account physical conditions that may affect the results of the set of spectral measurements.

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

As further shown in FIG6 , process 600 may include performing a first classification based on the results of the set of spectral measurements and a global classification model (block 620). For example, the control device 210 may perform the first classification based on the results and the global classification model. In some implementations, the control device 210 may receive the global classification model for use in performing the first classification. For example, the first control device 210 may generate a global classification model (e.g., using an SVM linear classifier and based on a set of spectral measurements performed on a training set, as described herein with respect to FIG4 ), and may provide the global classification model to the second control device 210 for use in performing the first classification of the unknown set. Additionally or alternatively, the control device 210 may generate a global classification model (e.g., using an SVM linear classifier and based on a set of spectral measurements performed on a training set, as described herein with respect to FIG4 ), and may utilize the global classification model for use in performing the first classification of the unknown set.

In some implementations, the control device 210 may assign a specific sample of the unknown group to a specific category in a set of categories of the global classification model when performing the first classification. For example, the control device 210 may determine, based on the global classification model, that a specific spectrum associated with the specific sample corresponds to a class of compounds (e.g., cellulose compounds, lactose compounds, caffeine compounds, etc.), and may assign the specific sample to a specific category. In some implementations, the control device 210 may assign the specific sample based on a confidence metric. For example, the control device 210 may determine, based on the global classification model, the probability that a specific spectrum is associated with each category of the global classification model. In this case, the control device 210 may assign the specific sample to the specific category based on the specific probability for the specific category exceeding the probability associated with other categories. In this way, the control device 210 determines a type of compound associated with the sample, thereby identifying the sample.

Additionally or alternatively, the control device 210 may determine another confidence metric associated with the first classification. For example, when the control device 210 assigns a particular sample to a particular category when performing the first classification, the control device 210 may determine the difference between the probability that the particular sample is associated with the particular category (e.g., referred to as the maximum probability) and the probability that the particular sample is associated with the next most likely particular category (e.g., referred to as the second maximum probability). In this way, the control device 210 determines the confidence associated with assigning the particular sample to a particular category rather than the next most likely category. When both the maximum probability and the second maximum probability are relatively high and relatively similar (e.g., the maximum probability is 48% and the second maximum probability is 47%, rather than the maximum probability being 48% and the second maximum probability being 4%), the control device 210 provides a better indication of the accuracy of the assignment by providing the difference between the maximum probability and the second maximum probability. In other words, in the first case where the maximum probability is 48% and the second maximum probability is 47%, the accuracy of the assignment to the most likely category is relatively lower than in the second case where the maximum probability is 48% and the second maximum probability is 4%, although the maximum probability is the same for both cases. Providing a measure of the difference between the maximum probability and the second maximum probability can distinguish between these two cases.

As further shown in FIG6 , process 600 may include generating a local classification model based on the first classification (block 630). For example, the control device 210 may generate the local classification model based on the first classification. The local classification model may involve an in-situ classification model generated based on a confidence metric associated with the first classification using an SVM classification technique (e.g., SVM-rbf, SVM linear, etc.; SVM based on probability values, SVM based on decision values, etc.; or similar techniques). For example, when a set of confidence metrics are determined for the spectrum of the sample based on a global classification model, the control device 210 may select a subset of categories of the global classification model based on the corresponding probabilities of the spectrum being associated with each category of the global classification model. In this case, the control device 210 may use the SVM classification technique and generate the local classification model based on the selected subset of categories.

In some implementations, a scaling preprocessing process may be performed. For example, to generate a local classification model, the control device 210 may perform an automatic scaling preprocessing process on spectra associated with a subset of the categories of the global classification model selected for the local classification model. In some implementations, the automatic scaling preprocessing process may be performed for another classification, such as the first classification, that uses the global classification model. In some implementations, another type of preprocessing process may be performed, such as a centering process, a transformation, etc.

In some implementations, the subset of categories may include a threshold number of categories associated with the highest corresponding confidence metrics. For example, the control device 210 may select ten categories of the global classification model based on the ten categories being associated with a higher corresponding probability (with which the spectrum of the sample is associated) than other categories of the global classification model, and may generate a local model based on these ten categories. In some implementations, the control device 210 may select the subset of categories based on the subset of categories that meet a threshold. For example, the control device 210 may select each category associated with a probability that meets a threshold. Additionally or alternatively, the control device 210 may select a threshold number of categories that each meet a threshold. For example, the control device 210 may select up to ten categories, assuming that each of the ten categories meets a minimum threshold probability. Additionally or alternatively, the control device 210 may select another number of categories (e.g., two categories, five categories, twenty categories, etc.).

In some implementations, the control device 210 can generate multiple local classification models. For example, the control device 210 can generate a first local classification model for a first spectrum of a first sample in the unknown group and a second local classification model for a second spectrum of a second sample in the unknown group. In this manner, the control device 210 can facilitate simultaneous classification of multiple unknown samples by using multiple local classification models to operate on the multiple unknown samples simultaneously.

In some implementations, the control device 210 may generate a quantification model based on performing a first classification using a 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 may use the first classification to select a class for the unknown sample and may select a local quantification model associated with the class of the unknown sample. In this manner, the control device 210 utilizes a hierarchical classification and quantification model to improve raw material identification and/or quantification thereof.

As further shown in FIG6 , process 600 may include performing a second classification based on the results of the set of spectral measurements and the local classification model (block 640). For example, the control device 210 may perform the second classification based on the results and the local classification model. In some implementations, the control device 210 may perform the second classification on a particular spectrum. For example, the control device 210 may assign the particular spectrum to a particular category based on the local classification model. In some implementations, the control device 210 may determine a set of confidence metrics associated with the particular spectrum and the local classification model. For example, the control device 210 may determine the probability that the particular spectrum is associated with each category of the local classification model, and may assign the particular spectrum (e.g., a particular sample associated with the particular spectrum) to a category with a higher probability than other categories of the local classification model. In this manner, the control device 210 identifies samples of an unknown group.

Additionally or alternatively, the control device 210 may determine another confidence metric associated with a particular spectrum and the local classification model. For example, when the control device 210 assigns a particular sample to a particular category when performing the second classification, the control device 210 may determine the difference between the probability that the particular sample is associated with the particular category (e.g., the maximum probability) and the probability that the particular sample is associated with the next most likely category (e.g., the second largest probability). In this manner, the control device 210 determines a confidence level associated with assigning a particular sample to a particular category rather than the next most likely category when performing the second classification based on the local classification model.

In some implementations, the control device 210 may perform multiple second classifications. For example, the control device 210 may perform a second classification of a first spectrum associated with a first sample based on a first local classification model, and may perform another second classification of a second spectrum associated with a second sample based on a second local classification model. In this way, the control device 210 facilitates simultaneous classification of multiple samples of an unknown group. In some implementations, the control device 210 may omit a portion of the samples in the unknown group from the second classification. For example, when the control device 210 determines a confidence metric for assigning a particular sample to a particular category based on a global classification model and the confidence metric meets a threshold, the control device 210 may omit the particular sample from the second classification. In this way, the control device 210 may reduce resource utilization relative to performing the second classification on all samples of the unknown group.

In some implementations, the control device 210 may perform quantification after performing the first classification (and/or after performing the second classification). For example, the control device 210 may select a local quantification model based on performing one or more classifications, and may perform quantification related to 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 body, where the plant body is associated with multiple quantification models (for example, 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 classifications to identify a specific quantification model. In this case, the control device 210 may determine that the plant is grown indoors in winter based on performing the set of classifications, and may select a quantification model related to the plant growing indoors in winter for use in determining the concentration of the specific chemical.

As further shown in FIG6 , process 600 may include providing information identifying a classification of an unknown group based on performing a second classification (block 650). For example, the control device 210 may provide information identifying a classification of a sample of an unknown group based on performing a second classification. In some implementations, the control device 210 may provide information identifying a specific category of a particular sample. For example, the control device 210 may provide information indicating that a specific spectrum associated with a particular sample is determined to be associated with a specific category, thereby identifying the sample. In some implementations, the control device 210 may provide information indicating a confidence metric associated with assigning a particular sample to a specific category. For example, the control device 210 may provide information indicating the probability of identifying a particular sample being associated with a specific category, the difference between the maximum probability and the second maximum probability for a particular sample, and the like. In this way, the control device 210 provides information indicating the likelihood that a particular spectrum is accurately assigned to a specific category.

In some implementations, the control device 210 provides information identifying the categories for multiple samples. For example, the control device 210 may provide information indicating that a first sample of an unknown group is associated with a first category and a second sample of the unknown group is associated with a second category. In this manner, the control device 210 provides simultaneous identification of multiple samples.

In some implementations, the control device 210 may provide quantification based on performing a set of classifications and quantifications. For example, based on identifying a local quantification model, the control device 210 may provide information identifying the concentration of a substance in an unknown sample for which a set of classifications is used to select a quantification model for determining the concentration of the substance.

In some implementations, the control device 210 may provide an output related to the classification of the sample. For example, the control device 210 may provide a binary output related to the classification of the unknown sample (e.g., a yes/no output) based on classifying the unknown sample into one of a first set of categories and a second set of categories, for which the first set of categories corresponds to a first binary output (e.g., yes) and the second set of categories corresponds to a second binary output (e.g., no). As an example, for a first set of categories (e.g., kosher meat, which may include kosher beef strip steak, kosher beef ribs, kosher chicken legs, kosher chicken breasts, etc.) and a second set of categories (e.g., non-kosher meat, which may include non-kosher beef ribs, non-kosher pork, non-kosher chicken wings, etc.), the control device 210 may provide an output of kosher or non-kosher based on classifying the unknown sample into a particular category of the first set of categories or the second set of categories. As another example, the control device 210 may utilize a set of categories associated with foods that are classified as halal or non-halal, and may provide an output indicating whether a sample corresponds to a halal category or a non-halal category (i.e., whether the animal from which the sample was obtained was slaughtered in a halal manner, without regard to whether other criteria for halal classification are met, such as religious certification, prayers during slaughter, etc.). In this manner, when identification of a particular category is not important to a user of the control device 210 (i.e., someone trying to determine whether a piece of meat is kosher rather than trying to determine the type of meat), the control device 210 may provide a classification with a greater likelihood of accuracy associated with providing identification of the particular category.

Although Figure 6 shows example blocks of process 600, in some implementations, process 600 may include additional blocks, fewer blocks, different blocks, or differently arranged blocks compared to those depicted in Figure 6. Additionally or alternatively, two or more blocks of process 600 may be performed in parallel.

Figures 7A and 7B are diagrams of an example implementation 700 related to prediction success rates associated with the example process 600 shown in Figure 6. Figures 7A and 7B illustrate example results of raw material identification using a hierarchical support vector machine (hier-SVM linear) based technique.

As shown in Fig. 7 A and by reference numeral 710, provide a group of confidence metrics for unknown group.For each sample of unknown group, control device 210 determines the probability that sample is relevant to each category of global classification model.For each sample of unknown group, maximum probability is compared with second maximum (next maximum) probability.As shown by reference numeral 712, the maximum probability range of unknown group is from about 5% to about 20%.As shown by reference numeral 714, the second maximum probability range of unknown group is from about 0% to about 5%.As shown by reference numeral 716, highlight the sample of unknown group that control device 210 incorrectly classifies based on global classification model (for example, 84 samples in 2645 samples in unknown group are incorrectly classified).

As shown in Fig. 7 A and by reference numeral 720, provide a group of confidence metrics for unknown group.For each sample of unknown group, control device 210 determines the probability that sample is relevant to each category of corresponding global classification model.For each sample of unknown group, maximum probability is compared with second maximum (next maximum) probability.As shown by reference numeral 722, the maximum probability range of unknown group is from about 50% to about 98%.As shown by reference numeral 724, the second maximum probability range of unknown group is from about 2% to about 45%.And, for each sample of unknown group except a sample (for its probability difference is about 8%, and yet for its correct classification being performed), the probability difference between maximum probability and second maximum probability is greater than about 0.33 (33%).Based on performing a group of classification, control device 210 correctly classifies all members of unknown group.

About Fig. 7 B, when the quantity of the sample in each category of classification model (for example, global classification model, local classification model etc.) fails to meet threshold value, control device 210 can determine the confidence measure and the relevant prediction accuracy that reduce when the sample of unknown group is assigned to category.As shown by reference numeral 730, when the quantity of the sample in each category does not meet threshold value, control device 210 is carrying out the first classification based on global classification model and based on one group of local classification model (for example, based on the SVM classifier local classification model of probability) after carrying out the second classification to unknown group, 128 samples in 4451 samples are mistakenly classified.As shown by reference numeral 740, when control device 210 carries out another first classification based on global classification model and based on another group of local classification model (for example, based on the SVM classifier local classification model of decision value) when carrying out another second classification, control device 210 is mistakenly classified with 1 sample in 4451 samples.In this way, control device 210 utilizes the SVM classifier based on decision value to improve classification accuracy relative to the SVM classifier based on probability.

As indicated above, Figures 7A and 7B are provided as examples only. Other examples are possible and may differ from what is described with respect to Figures 7A and 7B.

In this manner, the control device 210 performs RMID using the global classification model and the local classification model generated based on the global classification model.

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

Such implementations are described herein in conjunction with thresholds. As used herein, satisfying a threshold may refer to a value being greater than a threshold, more than a threshold, higher than a threshold, greater than or equal to a threshold, less than a threshold, less than a threshold, lower than a threshold, less than or equal to a threshold, equal to a threshold, etc.

It will be apparent that the systems and/or methods described herein may be implemented in various forms of hardware, firmware, or a combination of hardware and software. The actual dedicated control hardware or software code used to implement these systems and/or methods is not limiting of the implementation. Thus, the operation and behavior of the systems and/or methods are described herein without reference to specific software code—it being understood that software and hardware can be designed to implement the systems and/or methods based on the description herein.

Even if 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 listed 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 group.

The elements, actions or instructions used in this article should not be interpreted as critical or necessary unless explicitly described as such. In addition, as used in this article, the articles "a" and "an" are intended to include one or more items and can be used interchangeably with "one or more". In addition, as used in this article, the term "group" is intended to include one or more items (for example, related items, unrelated items, a combination of related items and unrelated items, etc.), and can be used interchangeably with "one or more". Where only one item is expected, the term "one" or similar language is used. In addition, as used in this article, the terms "has", "have", "having" etc. are defined as open terms. In addition, the phrase "based on" is intended to mean "based at least in part on", unless explicitly specified to the contrary.

Claims (14)

1. A sample identification device, comprising: A device for receiving a global classification model from another device. The global classification model is generated by the other device using support vector machine classifier technology and based on spectral measurements, and is distributed to multiple devices; The plurality of devices includes the sample identification device, and The global classification model includes a set of categories; A device for receiving information from the results of spectral measurements used to identify unknown samples; A means for performing a first classification of the unknown sample based on the results of the spectral measurements and the global classification model; Apparatus for generating a local classification model using support vector machine classifier technology and based on the first classification; The local classification model includes a subset of the set of categories; A means for performing a second classification of the unknown sample based on the results of the spectral measurements and the local classification model; and A means for providing information on the category associated with the unknown sample based on performing the second classification. 2. The sample identification device as described in claim 1, further comprising: A means for determining a set of corresponding probabilities associated with said set of categories. A specific probability in the set of corresponding probabilities indicates the likelihood that the unknown sample is associated with a specific category in the set of categories; and A means for selecting the subset of the set of categories based on the set of corresponding probabilities; and The apparatus for generating the local classification model includes: A means for generating the local classification model based on the subset of the set of categories. 3. The sample identification device as described in claim 1, further comprising: A device for performing an automatic scaling preprocessing procedure; and A means for performing at least one of the first classification or the second classification based on the execution of the automatic scaling preprocessing procedure. 4. The sample identification device as described in claim 1, The other device mentioned above is a control device associated with the first spectrometer; The global classification model is generated using one or more spectral measurements performed by the first spectrometer; The sample identification device further includes: A means for enabling the spectral measurement to be performed by a second spectrometer. The second spectrometer differs from the first spectrometer; and The apparatus for performing the first classification of the unknown sample includes: An apparatus for performing the first classification of the unknown sample based on the global classification model generated using the one or more spectral measurements performed by the first spectrometer and based on the results of spectral measurements performed by the second spectrometer. 5. The sample identification device of claim 1, wherein a set of categories of the global classification model corresponds to a set of compounds, and the categories are included in the set of categories; and The means for providing information for identifying the category includes: A means for providing information for identifying compounds in the group of compounds that correspond to the category. 6. The sample identification device of claim 1, wherein the means for performing the second classification comprises: A means for determining the relationship between the spectrum associated with the unknown sample and the category. The spectrum is identified by the result of performing the spectral measurement; and The means for providing information for identifying the category includes: A means for providing information for identifying a category based on determining the correlation between the spectrum associated with the unknown sample and the category. 7. The sample identification device as described in claim 1, The support vector machine classifier technique mentioned above is associated with at least one of the following: Kernel functions of radial basis function type, Kernel functions of linear function type Kernel functions of the S-type function type Kernel functions of polynomial function type, or Kernel functions of the exponential function type. 8. The sample identification device of claim 1, wherein the means for performing the second classification comprises: Apparatus for assigning the unknown sample to the category based on at least one of the following: probability value, or Decision value. 9. A sample identification method, comprising: The sample identification device receives the global classification model from another device. The global classification model is generated by the additional device and distributed to multiple devices; The plurality of devices includes the sample identification device, and The global classification model includes a set of categories; The sample identification device receives information about the results of spectral measurements performed by the first spectrometer to identify the unknown sample. Based on the results of the spectral measurements and the global classification model, the sample identification device performs a first classification of the unknown sample. The global classification model is generated using support vector machine classifier technology and a set of spectral measurements performed by a second spectrometer associated with the additional device. Based on the first classification, the sample identification device generates a local classification model. The local classification model utilizes support vector machine classifier technology, and The local classification model includes a subset of the categories in the set of categories of the global classification model; Based on the results of the spectral measurements and the local classification model, the sample identification device performs a second classification of the unknown sample; and Based on the information about the categories associated with the unknown sample provided by the sample identification device in the subset of the identification categories during the second classification. 10. The sample identification method of claim 9, wherein the kernel function associated with the support vector machine classifier technique includes at least one of the following: Kernel functions of radial basis function type, Kernel functions of linear function type Kernel functions of the S-type function type Kernel functions of polynomial function type, or Kernel functions of the exponential function type. 11. The sample identification method of claim 9, wherein performing the second classification includes: The unknown sample is assigned to the category of the subset of the category based on at least one of the following: The probability value associated with the category, or Decision values associated with the category. 12. The sample identification method as described in claim 9, wherein the first spectrometer is different from the second spectrometer. 13. The sample identification method as described in claim 9, further comprising: The global classification model is stored via a data structure. The execution of the first classification includes: The global classification model is obtained from the data structure; and The first classification is performed using the global classification model. 14. The sample identification method as described in claim 9, further comprising: Provides information to identify the confidence metric associated with the second classification. The confidence metric represents a measure of the confidence that the unknown sample is assigned to the category.