# US 2021/0241075 A1 · System for generating synthetic digital data of multiple sources > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 2021/0241075 A1 - Document ID: US20210241075A1 - Status: Published application - Kind: A1 - Application number: 16/778,929 - Filed: 2020-01-31 - Published: 2021-08-05 - Earliest claimed: 2020-01-31 - Assignees: Rohde & Schwarz GmbH & Co. KG - Applicants: Rohde & Schwarz GmbH & Co. KG - Inventors: Timo MAYER; Mikhail VOLIANSKII - Web page: https://patentgrep.com/patent/US20210241075A1 ## CPC - G06N 3/045: Combinations of networks - G06N 3/088: Non-supervised learning, e.g. competitive learning - G06F 11/3082: {the data filtering being achieved by aggregating or compressing the monitored data} - G06F 11/3089: {Monitoring arrangements determined by the means or processing involved in sensing the monitored data, e.g. interfaces, connectors, sensors, probes, agents (software debugging using additional hardware using a specific debug interface G06F11/3656; performance evaluation by tracing or monitoring G06F11/3466)} - G06N 3/0455: Auto-encoder networks; Encoder-decoder networks - G06N 3/0475: Generative networks - G06N 3/08: Learning methods - G06N 3/09: Supervised learning - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) ## Abstract The invention relates to a system for generating synthetic digital data, comprising: a receiver configured to receive at least one measured signal, in particular an RF signal or a sensor signal, a converter configured to convert the at least one measured signal to a digital dataset representing signal characteristics of the at least one measured signal, at least one trainable neural network encoder, wherein, during a training routine, the neural network encoder is configured to receive the digital dataset and to generate a compressed representation of the digital dataset, a processing unit configured to analyze the compressed representation and to detect a correlation between the digital dataset and the compressed representation, wherein the processing unit is configured to generate decoder input data based on the detected correlation, and a trained neural network decoder which is configured to receive the decoder input data and to generate synthetic digital data representing signal characteristics of the at least one measured signal based on the decoder input data. ## Claims ### Claim 1 A system for generating synthetic digital data, comprising: a receiver configured to receive at least one measured signal, in particular an RF signal or a sensor signal, a converter configured to convert the at least one measured signal to a digital dataset representing signal characteristics of the at least one measured signal, at least one trainable neural network encoder, wherein during a training routine the neural network encoder is configured to receive the digital dataset and to generate a compressed representation of the digital dataset, a processing unit configured to analyze the compressed representation and to detect a correlation between the digital dataset and the compressed representation, wherein the processing unit is configured to generate decoder input data based on the detected correlation, and a trained neural network decoder which is configured to receive the decoder input data and to generate synthetic digital data representing signal characteristics of the at least one measured signal based on the decoder input data. ### Claim 2 (depends on 1) The system of claim 1, wherein the processing unit is configured to generate the decoder input data by superimposing random or pseudorandom data with information based on the detected correlation, wherein the information based on the detected correlation represents constraints on the random or pseudorandom data. ### Claim 3 (depends on 1) The system of claim 1, wherein the digital dataset comprises at least two data segments, wherein the system comprises a plurality of trainable neural network encoders, wherein, during the training routine, each neural network encoder is configured to receive a data segment of the digital dataset and to generate a compressed representation of the respective data segment. ### Claim 4 (depends on 3) The system of claim 3, wherein the processing unit is configured to analyze the compressed representations of each data segment and to detect a correlation between each data segment and the corresponding compressed representation, in particular by comparing the compressed representations with each other. ### Claim 5 (depends on 4) The system of claim 4, wherein the processing unit is configured to generate the decoder input data based on the detected correlations for each data segment. ### Claim 6 A method for generating synthetic digital data, comprising the steps of: training at least one neural network encoder, by: a) receiving at least one measured signal, in particular an RF signal or a sensor signal, b) converting the at least one measured signal to a digital dataset representing signal characteristics of the at least one measured signal, and c) feeding said digital dataset to the at least one neural network encoder, wherein the neural network encoder is configured to generate a compressed representation of the digital dataset, analyzing the compressed representation to detect a correlation between the digital dataset and the compressed representation, generating decoder input data based on the detected correlation, feeding the decoder input data to a trained neural network decoder, wherein the trained neural network decoder is configured to generate synthetic digital data representing signal characteristics of the at least one measured signal based on the decoder input data. ### Claim 7 (depends on 6) The method of claim 6, wherein the decoder input data is generated by superimposing random or pseudorandom data with information based on the detected correlation, wherein the information based on the detected correlation represents constraints on the random or pseudorandom data. ### Claim 8 (depends on 6) The method of claim 6, wherein the digital dataset comprises at least two data segments, wherein a plurality of neural network encoders are trained, wherein during the training routine each neural network encoder is configured to receive a data segment of the digital dataset and to generate a compressed representation of the respective data segment. ### Claim 9 (depends on 8) The method of claim 8, wherein the step of detecting a correlation between the digital dataset and the compressed representation comprises detecting a correlation between each data segment and the corresponding compressed representation, in particular by comparing the compressed representations with each other. ### Claim 10 (depends on 9) The method of claim 9, wherein the decoder input data is generated based on the detected correlations for the at least two data segments. ### Claim 11 (depends on 7) The method of claim 7, wherein the at least one neural network encoder and the neural network decoder are further trained by: feeding the compressed representation of the digital dataset to the neural network decoder, wherein the neural network decoder is configured to generate a reconstruction of the digital dataset based on the compressed representation, comparing the reconstructed digital dataset generated by the neural network decoder to the corresponding digital dataset fed to the neural network encoder, determining a loss function, representing an accuracy loss between the digital dataset and the corresponding reconstructed digital dataset, and adapting the neural network encoder and/or the neural network decoder based on the loss function. ### Claim 12 (depends on 6) A test method for a physical entity, in particular a communication channel or a measurement device, comprising feeding the entity with the synthetic data generated according to the method of claim 6, and performing a real world measurement as to characteristics of the entity. ### Claim 13 (depends on 6) Use of synthetic digital data generated according to the method of claim 6 in a test method for a physical entity. ### Claim 14 (depends on 6) A computer program comprising a program code for performing the method of claim 6 when executed on a computer.