Adaptive Speech Enhancement Triggered by Voice Commands
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automatic speech recognition (ASR) systems in mobile devices face challenges in enhancing speech data quickly after waking up from a low-power standby mode due to the time required for adaptive noise reduction and beamforming algorithms to adapt to the acoustic environment, leading to poor recognition results.
Innovation Solution
A method that stores received speech data and uses a trigger phrase to train an adaptive speech enhancement block, allowing it to converge on optimal parameters before processing the data, ensuring enhanced speech output for immediate recognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If adaptive noise reduction and beamforming algorithms are used to enhance speech data, then speech recognition accuracy is improved, but the adaptation time causes delays in processing speech immediately after waking from standby mode
Solution Approach 1:
The system performs preliminary actions by continuously monitoring for trigger phrases during standby mode and pre-loading enhancement algorithms into memory before full processing is needed. When a trigger phrase is detected, the enhancement block is already prepared and can immediately process the speech data without requiring full adaptation from scratch, thus reducing the effective adaptation delay while maintaining recognition accuracy.
2Measurement precision
If adaptive speech enhancement algorithms are continuously running to improve recognition rates, then speech quality is enhanced, but power consumption increases significantly
Solution Approach 1:
The system implements periodic action by keeping the speech enhancement block in a low-power standby state during most periods, only activating it when a trigger phrase is detected. The system periodically checks for trigger phrases using minimal power, and only when needed does it fully activate the enhancement algorithms, thus maintaining recognition accuracy while dramatically reducing average power consumption compared to continuous operation.
Solution Approach 2:
The system uses the detected trigger phrase itself as the signal to activate the enhancement block, making the system self-regulating. The speech enhancement service activates itself automatically when needed (when trigger phrase is present) and deactivates when not needed, eliminating the need for external power management control while optimizing the balance between recognition accuracy and power consumption.
3Measurement precision
If the speech enhancement block waits for user speech to begin adaptation, then adaptation is based on actual speech data, but the enhancement is not available immediately when speech begins
Solution Approach 1:
The system performs preliminary actions by pre-loading the speech enhancement algorithms and parameters into memory during standby mode, before actual speech processing begins. When a trigger phrase is detected, the enhancement block is already prepared and can immediately start processing speech data without waiting for adaptation to complete, thus providing both immediate enhancement availability and quality adaptation based on actual speech data.
Data Source
AI summary
Received data representing speech is stored, and a trigger detection block detects a presence of data representing a trigger phrase in the received data. In response, a first part of the stored data representing at least a part of the trigger phrase is supplied to an adaptive speech enhancement block, which is trained on the first part of the stored data to derive adapted parameters for the speech enhancement block. A second part of the stored data, overlapping with the first part of the stored data, is supplied to the adaptive speech enhancement block operating with said adapted parameters, to form enhanced stored data. A second trigger phrase detection block detects the presence of data representing the trigger phrase in the enhanced stored data. In response, enhanced speech data are output from the speech enhancement block for further processing, such as speech recognition.