Low Power Speech Recognition Signal Selection Circuit
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing speech recognition systems in portable devices face challenges with high power consumption due to complex algorithms and high-performance analog-to-digital converters, making it impractical to keep them continually active with limited battery capacity.
Innovation Solution
A speech recognition system that includes an input signal buffer, noise reduction block, and a selection circuit to direct signals to a speech recognition engine, using a multi-phase process for hands-free operation with low power consumption, where only the necessary components are actively powered during speech detection and recognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high-performance analog-to-digital converters with feedback loops are used to ensure accurate speech signal digitisation, then measurement precision is improved, but use of energy increases
Solution Approach 1:
The audio processing system is divided into multiple functional stages: a first stage performs simple analog signal buffering and basic noise reduction with low power consumption, while a second stage performs high-precision analog-to-digital conversion only when needed. This segmentation allows the system to maintain measurement precision for speech recognition while significantly reducing average power consumption by keeping the high-precision converter inactive during most operations.
Solution Approach 2:
The system performs preliminary noise reduction and signal buffering using low-power circuitry before the high-power analog-to-digital converter is activated. This preliminary processing prepares the signal for accurate digitisation while minimizing the duration that the high-power components are active, thus resolving the contradiction between precision and energy consumption.
2Measurement precision
If complex algorithms are implemented for speech recognition and speaker verification, then measurement precision is improved, but use of energy increases
Solution Approach 1:
The speech processing algorithm is segmented into multiple stages: an initial low-power stage that performs basic audio buffering and noise reduction, followed by a higher-power stage that performs accurate speech recognition and speaker verification only when the pass phrase is detected. This segmentation enables the system to maintain high measurement precision for speech recognition while reducing average power consumption by keeping complex algorithms inactive during most time periods.
Solution Approach 2:
The system uses periodic monitoring at low power to detect pass phrases, then activates complex speech recognition algorithms periodically only when necessary. This periodic activation pattern allows the system to maintain high speech recognition accuracy when needed while minimizing the cumulative energy consumption from running complex algorithms continuously.
3Ease of operation
If hands-free operation is implemented with continuous listening capability, then ease of operation is improved, but use of energy increases
Solution Approach 1:
The system segments audio processing into a continuously active low-power first stage that performs basic signal buffering and noise reduction, and a second stage that performs hands-free speech recognition only when pass phrases are detected. This segmentation enables hands-free operation during speech recognition while minimizing power consumption by keeping the second stage inactive during most operations.
Solution Approach 2:
The system dynamically adjusts its operational state based on detected conditions: it maintains low-power continuous listening for basic audio capture, then transitions to high-power hands-free speech recognition mode only when pass phrases are detected. This dynamic state change allows the system to provide hands-free operation when needed while significantly reducing average power consumption during idle periods.
Data Source
AI summary
A speech recognition system comprises: an input, for receiving an input signal from at least one microphone; a first buffer, for storing the input signal; a noise reduction block, for receiving the input signal and generating a noise reduced input signal; a speech recognition engine, for receiving either the input signal output from the first buffer or the noise reduced input signal from the noise reduction block; and a selection circuit for directing either the input signal output from the first buffer or the noise reduced input signal from the noise reduction block to the speech recognition engine.


