Biological Signal Conversion Using 160 Hz Silent Speech Sensing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing information conversion systems face challenges in achieving sufficient recognition accuracy due to low data sampling frequencies, particularly in silent speech recognition, which can result in inadequate conversion of biological signals into text or speech information.
Innovation Solution
The system employs a biological information detection unit comprising acceleration and angular velocity sensors with a sampling frequency of at least 160 Hz, utilizing a trained model based on Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN) to convert biological signals into text or speech information, distinguishing between silent and vocalizing states.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a low sampling frequency (e.g., 58.3 Hz) is used for biological signal detection, then the device complexity and power consumption are reduced, but the conversion accuracy and recognition precision deteriorate
Solution Approach 1:
The patent changes the sampling frequency parameter from conventional low values (58.3 Hz) to a higher value (160 Hz or more) to improve measurement precision. This parameter change enables the system to capture sufficient temporal details of biological signals for accurate silent speech recognition, while the use of CNN and RNN models efficiently processes the increased data rate without proportionally increasing device complexity
2Measurement precision
If a high sampling frequency (e.g., 160 Hz) is used for biological signal detection, then the conversion accuracy and recognition precision are improved, but the data processing load and computational requirements increase
Solution Approach 1:
The patent replaces traditional mechanical signal processing methods with deep learning models (CNN and RNN) that can efficiently process high-frequency biological signals. The CNN model extracts spatial features from the high-rate acceleration and angular velocity data, while the RNN model captures temporal dependencies, achieving accurate phoneme recognition without requiring excessive computational power for manual feature engineering and processing
3Reliability
If acceleration and angular velocity sensors with sampling frequency ≥160 Hz are used, then silent speech recognition accuracy is improved, but the device cost and complexity increase
Solution Approach 1:
The patent makes the biological information detection unit multi-functional by using acceleration and angular velocity sensors not only for silent speech recognition but also for distinguishing between silent and vocalizing states. This universal use of the same sensors across multiple functions reduces the need for additional specialized sensors, thereby limiting the increase in device complexity while improving overall system reliability
Data Source
AI summary
There is provided an information conversion system having an improved conversion accuracy of converting biological information into text information or speech information.