Audio Endpoint Detection via Dynamic Mute Threshold Adjustment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing speech assessment systems require manual user input for audio record control, leading to complex operations and inefficient endpoint detection, where normal pronunciation pauses are misjudged as the end of the record, resulting in user voice truncation and prolonged waiting times.
Innovation Solution
A method and system that automatically detect the endpoint of an audio record by presetting a mute duration threshold, determining an acoustic model for the text endpoint, and updating the threshold to a shorter value when the acoustic model of the current frame matches the endpoint model, allowing for earlier termination of the recording.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the time threshold is set to be shorter to detect endpoint faster, then the identifying efficiency is improved, but normal pronunciation pauses are misjudged as endpoint causing user voice truncation
Solution Approach 1:
The patent dynamically adjusts the time threshold parameter based on acoustic model matching results. When the acoustic model of the current frame matches the endpoint model, the system switches to a shorter time threshold for faster detection. When there is no match, it uses a longer threshold to avoid misjudging pronunciation pauses. This dynamic parameter adjustment resolves the contradiction between detection speed and accuracy.
Solution Approach 2:
The system transitions from a static time threshold approach to a dynamic one where the threshold changes based on real-time acoustic analysis. The acoustic model matching mechanism enables the system to adaptively select appropriate threshold values, making the endpoint detection both fast and reliable by responding to the actual speech characteristics.
2Reliability
If the time threshold is set to be larger to avoid misjudging pronunciation pauses, then the detection accuracy is improved, but the waiting time for users is prolonged
Solution Approach 1:
The system changes the time threshold parameter dynamically based on acoustic model matching. When the acoustic model indicates an endpoint is likely, the system uses a shorter threshold to reduce waiting time. When the acoustic model suggests continued speech, it uses a longer threshold to ensure accuracy. This resolves the contradiction between accuracy and user waiting time.
Solution Approach 2:
The system performs preliminary acoustic model matching before final endpoint determination. By analyzing the acoustic characteristics in advance and comparing them with the endpoint model, the system can predict whether a shorter or longer threshold should be used, thereby optimizing both accuracy and response time before the actual endpoint decision is made.
3Measurement precision
If manual user input is required for audio record control, then the control precision is improved, but the operation complexity increases
Solution Approach 1:
The system performs automatic endpoint detection using acoustic model matching, eliminating the need for manual user input to control the audio record. The system independently analyzes the speech signal, compares it with the endpoint model, and automatically determines when to stop recording. This maintains precision while significantly improving ease of operation by removing manual intervention requirements.
Solution Approach 2:
The patent replaces the mechanical manual control system with an automated acoustic analysis system. Instead of relying on user clicks or inputs, the system uses acoustic model matching and signal processing to automatically detect endpoints. This substitution maintains control precision through sophisticated analysis while eliminating operation complexity by removing manual interaction.
Data Source
AI summary
A method and system for endpoint automatic detection of audio record is provided. The method comprises the following steps: acquiring a audio record text and affirming the text endpoint acoustic model for the audio record text; starting acquiring the audio record data of each frame in turn from the audio record start frame in the audio record data; affirming the characteristics acoustic model of the decoding optimal path for the acquired current frame of the audio record data; comparing the characteristics acoustic model of the decoding optimal path acquired from the current frame of the audio record data with the endpoint acoustic model to determine if they are the same; if yes, updating a mute duration threshold with a second time threshold, wherein the second time threshold is less than a first time threshold. This method can improve the recognizing efficiency of the audio record endpoint.


