EMG Inner Speech Detection Using ML Artifact Rejection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional EMG systems face challenges in accurately detecting silent speech due to external interference from involuntary gestures, requiring lengthy training, and being affected by individual differences in speech production.
Innovation Solution
A system using machine learning (ML) models to detect inner speech by processing EMG signals from a wearable device, allowing users to interact with applications and XR devices without overt physical movement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional EMG systems are used to detect silent speech, then speech detection capability is provided, but detection accuracy deteriorates due to external interference from involuntary gestures
Solution Approach 1:
The patent segments the EMG signal processing into multiple independent components: raw signal acquisition, artifact detection module, machine learning classification, and speech output. By dividing the detection process into discrete segments, the system can identify and exclude involuntary gesture artifacts while preserving genuine speech signals, thereby improving detection accuracy despite the presence of interfering gestures.
Solution Approach 2:
The patent introduces machine learning models as an intermediary between raw EMG signals and speech detection decisions. This intermediary layer processes the signals, learns to distinguish between speech-related muscle activity and involuntary gestures, and makes informed classification decisions. The ML model acts as a mediator that filters out harmful gesture interference while preserving legitimate speech detection.
2Reliability
If conventional EMG systems implement artifact rejection mechanisms, then detection reliability improves, but system complexity increases
Solution Approach 1:
The patent implements a self-service artifact rejection system where the machine learning model automatically learns to distinguish speech from gestures during training and autonomously performs classification during operation. The system serves itself by continuously improving its discrimination capabilities without requiring manual intervention or complex external artifact rejection hardware, thereby maintaining high reliability while controlling system complexity.
Solution Approach 2:
The patent changes the operational parameters of the EMG system by transitioning from traditional threshold-based detection to machine learning-based probabilistic classification. This parameter change allows the system to adapt to different users and conditions, improving reliability through learned patterns while the modular ML implementation keeps complexity manageable through standardized algorithms.
3Measurement precision
If machine learning models are used to process EMG signals, then speech detection accuracy improves, but computational resource consumption increases
Solution Approach 1:
The patent applies partial action by using machine learning models selectively rather than continuously. The system employs ML classification primarily during training phases and critical detection moments, while using simpler threshold-based methods or pre-computed features during routine operation. This partial application of computationally intensive processing maintains high detection accuracy when needed while reducing overall resource consumption during normal operation.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The system enhances the accuracy and efficiency of silent speech detection, reduces resource consumption, and improves user interaction with electronic devices by enabling noise-free and detectable inner speech.
Implementation Method 1
EMG speech systems and to interaction applications and/or extended reality (XR) devices... electromyograph (EMG) speech systems... detect electrical signals associated with muscle activity
Data Source
AI summary
Methods and systems are disclosed for training a machine learning (ML) model to detect inner speech. The system collects, by an electromyograph (EMG) communication device used by a user, a first set of EMG signals over a first time interval. The system generates a first plurality of features based on the first set of EMG signals and generates a first probability associated with presence of inner speech by processing the first plurality of features with a machine learning (ML) model. The system compares the first probability generated by the ML model to a specified threshold and detects presence of the inner speech of the user in response to determining that the first probability generated by the ML model transgresses the specified threshold.