Myoelectric Speech Recognition via Bone Conduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional speech recognition systems struggle with recognizing barely audible or whispered speech, and they perform poorly in noisy environments, leading to confidentiality and disturbance issues, as well as limitations in robustness and mobility.
Innovation Solution
A myoelectric-based processing method that captures and amplifies muscle contraction signals to convert them into digital signals for automatic speech recognition, allowing for silent speech recognition and translation without the need for audible vocal effort, using surface electromyography to interpret articulatory muscle activity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional speech recognition systems use air transmission microphones to capture speech, then they can recognize audible speech in quiet environments, but they perform poorly in recognizing barely audible or whispered speech and cannot recognize silently mouthed speech
Solution Approach 1:
The patent introduces an intermediary mechanism (bone conduction pathway) to transfer speech signals from the speaker's vocal apparatus directly to the microphone, bypassing the air transmission medium. This allows the system to capture both audible and silent speech with high accuracy, resolving the contradiction between recognition precision and adaptability to different speech types.
Solution Approach 2:
The patent replaces the acoustic field-based detection system with a mechanical vibration-based detection system. By using bone conduction to transmit mechanical vibrations directly to the microphone, the system can detect speech signals that are imperceptible through air transmission, thereby improving both measurement precision and adaptability to whispered and silent speech.
2Reliability
If speech recognition systems rely on audible speech transmission through air, then they can capture speech signals, but they cannot ensure confidentiality and may be disturbed by surrounding noise
Solution Approach 1:
The patent extracts the speech signal detection from the noisy air transmission medium and relocates it to the bone conduction pathway. By taking out the speech capture process from the external acoustic environment, the system achieves confidentiality and immunity to environmental noise, as bone-conducted vibrations are not affected by surrounding sounds.
3Reliability
If conventional systems use directional microphones or microphone arrays to improve noise rejection, then speech recognition in noisy environments may be improved, but the systems become less mobile and more expensive
Solution Approach 1:
The patent replaces the complex acoustic field-based microphone array system with a simple mechanical bone conduction microphone. This substitution eliminates the need for multiple microphones and complex signal processing algorithms, achieving noise-resistant speech recognition with a single, simple, and mobile device.
4Adaptability or versatility
If speech recognition systems require audible vocal effort, then they can capture speech signals, but they cannot recognize silently mouthed speech or speech with very low vocal effort
Solution Approach 1:
The patent introduces bone conduction as an intermediary pathway that directly couples the vocal apparatus to the detection sensor. This intermediary mechanism preserves the mechanical vibrations of speech production regardless of vocal effort level, enabling the system to detect silently mouthed speech, whispered speech, and normal speech with equal precision.
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
Enables robust and confidential speech recognition and translation in various environments, including noisy conditions, without disturbing others, and can be used for large vocabulary tasks across languages and speaking styles.
Implementation Method 1
capturing a myoelectric signal from a user using at least one electrode, wherein the electrode converts an ionic current generated by muscle contraction into an electric current
Data Source
AI summary
A method for myoelectric-based processing of speech. The method includes capturing a myoelectric signal from a user using at least one electrode, wherein the electrode converts an ionic current generated by muscle contraction into an electric current. The method also includes amplifying the electric current, filtering the amplified electric current, and converting the filtered electric current into a digital signal. The method further includes transmitting the myoelectric signal to a digital device, transforming the digital signal into a written representation using an automatic speech recognition method, and generating an audible output from the written representation using a speech synthesis method.