Methods and systems for real-time electronic verification of content with varying features in data-sparse computer environments
The systems and methods provide a machine learning model that can exploit long time dependency for time-series sequences, perform end-to-end learning of dimension reduction and clustering, or train on long time-series sequences with low computation complexity. For example, the methods and systems use a novel, unsupervised temporal representation learning model. The model may generate cluster-specific temporal representations for long-history time series sequences and may integrate temporal reconstruction and a clustering objective into a joint end-to-end model.
1. A system for real-time electronic signature verification in a data-sparse computer environment, the system comprising:
cloud-based storage circuitry configured to store a first image, wherein the first image comprises a first signature, wherein the first signature is an authenticated signature;
control circuitry configured to:
determine a first sequence, wherein the first sequence comprises a sequence of intensity for both vertical and horizontal concatenations of pixels composing the first image;
extract a first dynamic behavior of the first sequence, wherein extracting the first dynamic behavior of the first sequence comprises determining a first optimal stochastic transition matrix (OSTM) that represents the first dynamic behavior;
receive a second image, wherein the second image comprises a second signature, wherein the second signature requires electronic verification;
determine a second sequence, wherein the second sequence comprises a sequence of intensity for both vertical and horizontal concatenations of pixels composing the second image;
extract a second dynamic behavior of the second sequence, wherein extracting the second dynamic behavior of the second sequence comprises determining a second optimal stochastic transition matrix (OSTM) that represents the second dynamic behavior;
determine an authenticity of the second signature based on comparing the first dynamic behavior to the second dynamic behavior;
select a recommendation from a plurality of recommendations based on the authenticity; and
input/output circuitry configured to generate for display, on a user interface, the recommendation.
2. A method for real-time electronic signature verification in data-sparse computer environment, the method comprising:
receiving a first image, wherein the first image comprises a first signature;
determining a first sequence, wherein the first sequence comprises a sequence of intensity of pixels composing the first image;
extracting a first dynamic behavior of the first sequence, wherein extracting the first dynamic behavior of the first sequence comprises determining a first optimal stochastic transition matrix (OSTM) that represents the first dynamic behavior;
receiving a second image, wherein the second image comprises a second signature;
determining a second sequence, wherein the second sequence comprises a sequence of intensity of pixels composing the second image;
extracting a second dynamic behavior of the second sequence; and
generating for display, on a user interface, a recommendation based on comparing the first dynamic behavior to the second dynamic behavior.
3. The method of claim 2 , wherein generating for display the recommendation based on comparing the first dynamic behavior to the second dynamic behavior, further comprises:
determining an authenticity of the second signature based on comparing the first dynamic behavior to the second dynamic behavior, wherein the first signature is an authenticated signature, and wherein the second signature requires electronic verification; and
selecting the recommendation from a plurality of recommendations based on the authenticity.
4. The method of claim 3 , wherein determining the authenticity of the second signature based on comparing the first dynamic behavior to the second dynamic behavior comprises:
determining a deviation between the first dynamic behavior to the second dynamic behavior; and
comparing the deviation to a threshold deviation.
5. The method of claim 4 , wherein the deviation is based on a Euclidean distance.
6. The method of claim 2 , wherein generating for display the recommendation based on comparing the first dynamic behavior to the second dynamic behavior, further comprises:
generating a feature input based on the second dynamic behavior;
inputting the feature input into a machine learning model to determine a classification of the second dynamic behavior, wherein the machine learning model is trained based on the first dynamic behavior;
receiving an output from the machine learning model indicating the classification of the second dynamic behavior; and
selecting the recommendation from a plurality of recommendations based on the output.
7. The method of claim 2 , wherein the first sequence comprises both vertical and horizontal concatenations of pixels composing the first image.
8. The method of claim 2 , wherein determining the OSTM comprises:
partitioning the first sequence based on equidistance division on values in the first sequence;
determining a third sequence by mapping each sample of the first sequence into its corresponding partition, wherein the third sequence represents a sequence of states;
determining a stochastic transition matrix by estimating a probability of transition between the states; and
determining a number of partitions for determining the OSTM by determining a maximum entropy.
9. The method of claim 2 , wherein determining the OSTM comprises:
partitioning the first sequence based on k-means clustering on values in the first sequence;
determining a third sequence by mapping each sample of the first sequence into its corresponding partition, wherein the third sequence represents a sequence of states;
determining a stochastic transition matrix by estimating a probability of transition between states of the sequence of states; and
determining a number of partitions for determining the OSTM by determining a maximum entropy.
10. The method of claim 9 , wherein the number of partitions is further based on a stopping condition for updating the number of partitions.
11. A non-transitory, computer-readable medium for real-time electronic signature verification in a data-sparse computer environment, comprising instructions that, when executed by one or more processors, cause operations comprising:
receiving a first image, wherein the first image comprises a first signature;
determining a first sequence, wherein the first sequence comprises a sequence of intensity of pixels composing the first image;
extracting a first dynamic behavior of the first sequence, wherein extracting the first dynamic behavior of the first sequence comprises determining a first optimal stochastic transition matrix (OSTM) that represents the first dynamic behavior;
receiving a second image, wherein the second image comprises a second signature;
determining a second sequence, wherein the second sequence comprises a sequence of intensity of pixels composing the second image;
extracting a second dynamic behavior of the second sequence; and
generating for display, on a user interface, a recommendation based on comparing the first dynamic behavior to the second dynamic behavior.
12. The non-transitory, computer-readable medium of claim 11 , wherein generating for display the recommendation based on comparing the first dynamic behavior to the second dynamic behavior, further comprises:
determining an authenticity of the second signature based on comparing the first dynamic behavior to the second dynamic behavior, wherein the first signature is an authenticated signature, and wherein the second signature requires electronic verification; and
selecting the recommendation from a plurality of recommendations based on the authenticity.
13. The non-transitory, computer-readable medium of claim 12 , wherein determining the authenticity of the second signature based on comparing the first dynamic behavior to the second dynamic behavior comprises:
determining a deviation between the first dynamic behavior to the second dynamic behavior; and
comparing the deviation to a threshold deviation.
14. The non-transitory, computer-readable medium of claim 11 , wherein generating for display the recommendation based on comparing the first dynamic behavior to the second dynamic behavior, further comprises:
generating a feature input based on the second dynamic behavior;
inputting the feature input into a machine learning model to determine a classification of the second dynamic behavior, wherein the machine learning model is trained based on the first dynamic behavior;
receiving an output from the machine learning model indicating the classification of the second dynamic behavior; and
selecting the recommendation from a plurality of recommendations based on the output.
15. The non-transitory, computer-readable medium of claim 11 , wherein the first sequence comprises both vertical and horizontal concatenations of pixels composing the first image.
16. The non-transitory, computer-readable medium of claim 11 , wherein determining the OSTM comprises:
partitioning the first sequence based on equidistance division on values in the first sequence;
determining a third sequence by mapping each sample of the first sequence into its corresponding partition, wherein the third sequence represents a sequence of states;
determining a stochastic transition matrix by estimating a probability of transition between the states; and
determining a number of partitions for determining the OSTM by determining a maximum entropy.
17. The non-transitory, computer-readable medium of claim 11 , wherein the determining the OSTM comprises:
partitioning the first sequence based on k-means clustering on values in the first sequence;
determining a third sequence by mapping each sample of the first sequence into its corresponding partition, wherein the third sequence represents a sequence of states;
determining a stochastic transition matrix by estimating a probability of transition between states of the sequence of states; and
determining a number of partitions for determining the OSTM by determining a maximum entropy.
18. The non-transitory, computer-readable medium of claim 17 , wherein the number of partitions is further based on a stopping condition for updating the number of partitions.