Speech Overload Detection in ASR Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing communication systems with intelligent software agents struggle to accurately detect speech intervals and interpret user inputs, leading to disjointed conversations due to false recognition of speech, especially in noisy environments or when users speak too long or too fast, which affects the realism and efficiency of interactions.
Innovation Solution
The system employs an Automatic Speech Recognition (ASR) system to identify individual words and detect speech overload conditions, providing real-time notifications to users when their input becomes uninterpretable, allowing for more efficient and natural interactions by interrupting the user and adjusting response delivery based on speech activity and inactivity intervals.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the ASR system continuously processes voice input during speech intervals, then the system can identify words in real-time, but the system may become uninterpretable when users speak too long or too fast (speech overload)
Solution Approach 1:
The system performs preliminary detection of speech overload conditions during the speech interval by monitoring the number of identified words and their identification rates. When a threshold is exceeded, the system proactively notifies the user before complete uninterpretability occurs, allowing preventive action while maintaining continuous processing capability.
Solution Approach 2:
The system implements feedback by notifying the user when speech overload conditions are detected. This feedback loop allows the user to adjust their speaking rate or pause, enabling the ASR system to maintain reliable interpretation by resetting the word count counter after notification, thus balancing continuous processing with interpretation accuracy.
2Reliability
If the system notifies the user during speech interval, then the user can be alerted before speech becomes uninterpretable, but the notification may interrupt the user's speech flow
Solution Approach 1:
The system applies partial action by notifying the user only when speech overload thresholds are exceeded, rather than continuously interrupting. The notification is triggered selectively based on monitored parameters (word count, identification rate), providing just enough intervention to maintain reliability while minimizing disruption to natural speech flow.
3Measurement precision
If the system monitors word identification rate and count continuously, then speech overload can be detected accurately, but the system complexity increases
Solution Approach 1:
The ASR system's word identification functionality serves multiple purposes: it not only transcribes speech but also simultaneously monitors word count and identification rates for overload detection. This multi-functionality eliminates the need for separate monitoring components, achieving precise speech overload detection while minimizing system complexity through resource sharing.
Data Source
AI summary
A computer system comprises an input configured to receive voice input from a user, the voice input having speech intervals separated by non-speech intervals; an ASR system configured to identify individual words in the voice input during speech intervals of the voice input, and store the identified words in memory; a speech overload detection module configured to detect at a time during a speech interval of the voice input a speech overload condition; and a notification module configured to output to the user, in response to said to detection, a notification of the speech overload condition.