Hardware Voice Activity Detection with Context-Aware Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional voice activity detection systems face challenges in achieving high accuracy while minimizing power consumption, particularly in mobile and IoT devices, due to their always-on nature, which leads to battery life reduction and increased resource usage.
Innovation Solution
A context-aware hardware-based voice activity detection system utilizing a fast Fourier transform (FFT) for efficient feature generation and a hardware-based machine learning model with context-specific parameters to improve accuracy and reduce power consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If voice activity detection is implemented as always-on function to maximize utility, then detection accuracy is improved, but power consumption increases
Solution Approach 1:
The system segments the voice activity detection process into two distinct stages: a low-power initial detection stage that continuously monitors audio data, and a high-accuracy verification stage that activates only when the initial detection indicates potential voice activity. This segmentation allows the system to maintain high overall detection accuracy while minimizing power consumption by keeping the high-power components dormant most of the time.
Solution Approach 2:
The system implements periodic action by switching between different operational modes based on detected conditions. The low-power detector operates continuously at a basic level, while the high-accuracy detector operates periodically only when triggered by the initial detection. This periodic activation of high-power components resolves the contradiction between always-on accuracy and power consumption.
2Use of energy by moving object
If voice activity detection is implemented as low-accuracy, low-power function in mobile devices, then power consumption is reduced, but detection accuracy deteriorates
Solution Approach 1:
The system segments the detection task between a always-on low-power detector and an on-demand high-accuracy detector. The low-power detector handles routine monitoring with acceptable accuracy, while the high-accuracy detector is activated only when needed, ensuring that mobile devices achieve high detection accuracy without continuously consuming high power.
Solution Approach 2:
The low-power detector serves as an intermediary component that filters audio data before it reaches the high-accuracy detector. By pre-processing and filtering potential voice activities, it reduces the workload on the high-accuracy detector while maintaining overall system accuracy, thus resolving the contradiction between power consumption and detection accuracy in mobile devices.
3Measurement precision
If high-accuracy voice activity detection is implemented, then detection precision is improved, but device complexity increases
Solution Approach 1:
The system segments the detection architecture into hierarchical layers with increasing complexity. The first layer uses simple, low-complexity algorithms for continuous monitoring, while the second layer employs more complex high-accuracy algorithms only when triggered. This segmentation maintains high detection precision while managing device complexity by keeping complex components dormant and inactive most of the time.
Data Source
AI summary
Certain aspects of the present disclosure provide a method for performing voice activity detection, including: receiving audio data from an audio source of an electronic device; generating a plurality of model input features using a hardware-based feature generator based on the received audio data; providing the plurality of model input features to a hardware-based voice activity detection model; receiving an output value from the hardware-based voice activity detection model; and determining a presence of voice activity in the audio data based on the output value.