Systems and methods for transmitting and receiving data using machine learning classification
Inventors
Farsad, Nariman • Goldsmith, Andrea
Assignees
Leland Stanford Junior University
Publication Number
US-12314827-B2
Publication Date
2025-05-27
Expiration Date
2038-02-14
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Abstract
Systems and methods detect messages in transmitted and received data in communication systems in accordance with embodiments of the invention. In one embodiment, a communication system controller includes a processor, a memory, and a receiver, wherein the processor obtains a transmission signal using the receiver, extracts features in the transmission signal, and detects a message in the transmission signal based on the extracted features using a machine learning classifier.
Core Innovation
The invention relates to systems and methods for transmitting and receiving data in communication systems using machine learning classification. In these systems, a communication system controller that includes a processor, a memory, and a receiver is used to obtain a transmission signal, extract features from the signal, and then detect messages within the signal using a machine learning classifier. The classifier can be trained or retrained based on training data, including classified signals, to improve detection accuracy.
The problem addressed by the invention lies in the challenge of reliably detecting and classifying messages in communication signals, especially when the characteristics of the communication channel are not known or are difficult to measure. Existing systems may rely on prior knowledge of the channel properties, which is not always available or feasible to obtain, thus limiting the reliability and throughput of communication systems.
The core innovation introduces a machine learning-based approach where features are extracted from transmission signals and used by classifiers such as recurrent neural networks, support vector machines, and others to detect messages. The classifiers are trained using supervised learning with known data, and can be further retrained when detection fails, allowing continuous adaptation to varying channel conditions and signal properties. Feature extraction may consider signal attributes such as phase, magnitude, rate of change, and average power, and the method can be used across multiple communication channel types, including electrical, optical, electromagnetic, acoustic, and chemical channels.
Claims Coverage
The patent contains two independent claims, each introducing central inventive features concerning the use of a machine learning classifier for detecting message content in communication systems.
Communication system controller implementing machine learning classification for message detection
A communication system controller comprising: - A processor, receiver, and memory containing instructions; - The processor obtains a transmission signal via the receiver, the signal being encoded with a message and transmitted through a communication channel; - The processor extracts features from the transmission signal; - The system detects the message content based on the extracted features using a machine learning classifier; - The machine learning classifier is trained by receiving training messages, comparing detected and known values, and updating based on the difference; - The system further trains the machine learning classifier using the classified transmission signal.
Method for detecting message content in signals using machine learning classification
A method comprising: - Obtaining a transmission signal using a communication system controller that includes processor, memory, and receiver, with the signal encoded with a message and received via a communication channel; - Extracting features from the transmission signal using the controller; - Detecting the message content based on extracted features using a machine learning classifier trained by receiving a training message, comparing detected and known values, and updating based on the difference; - Further training the machine learning classifier using the classified transmission signal.
These inventive features collectively cover both system and method aspects for detecting and classifying messages in received signals using a machine learning classifier that is trained and retrained based on comparative analysis of detected and known message content, with applications to various types of communication channels.
Stated Advantages
Enables reliable message detection in communication systems without prior knowledge of channel characteristics.
Allows adaptation and continuous improvement of classification accuracy through retraining of the machine learning classifier using new or misclassified data.
Supports extraction and selection of relevant features from transmission signals to improve the efficiency and accuracy of the classifier.
Applicable to a wide range of communication channels, including electrical, optical, electromagnetic, acoustic, and chemical.
Documented Applications
Communication systems transmitting and receiving messages over channels such as twisted-pair wires, coaxial cables, fiber optic cables, tubes for chemicals, microwave, satellite, radio, and infrared.
Use in chemical communication systems transmitting messages via acid and base signals, including pH-based chemical signaling.
Applications involving detection and classification of messages in signals for telegraph, radio, telephone, cellular, acoustic communication, and cable systems.
Interested in licensing this patent?