Voice Assistant Acoustic Testing via Vocoder Parameter Modification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Voice assistant devices face challenges in accurately interpreting voice commands in noisy environments and varying voice types, leading to inadequate testing and increased error rates, which negatively impact user experience.
Innovation Solution
A computer-implemented method that generates modified user utterances with varying acoustic characteristics using vocoder parameters, tests voice assistant devices in diverse environments, and fine-tunes these parameters to improve noise and modulation handling, thereby enhancing response accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If voice assistant devices are tested with standard voice commands only, then testing process is simple, but the device fails to accurately interpret voice commands in noisy environments and with varying voice types
Solution Approach 1:
The patent applies parameter changes by modifying acoustic characteristics of test utterances through vocoder parameters. The system generates modified user utterances with varied acoustic properties (pitch, timbre, modulation) to test the voice assistant under diverse conditions. This resolves the contradiction by enabling comprehensive reliability testing without manually creating complex test scenarios, as the parameter transformation is automated through the vocoder.
Solution Approach 2:
The patent uses copying by creating synthetic modified versions of original user utterances. Instead of recording numerous real-world voice samples in different conditions, the system generates acoustic copies of standard test phrases with varied characteristics using vocoders. This allows comprehensive testing coverage while maintaining a simple base test set, resolving the contradiction between testing thoroughness and process simplicity.
2Measurement precision
If voice assistant devices are not tested with varying voice types and modulations, then testing is easier, but the device cannot discern naturally modulated words or phrases from standard words
Solution Approach 1:
The patent systematically varies acoustic parameters including pitch contour, timbre, and modulation depth through vocoder control to generate test utterances with different voice characteristics. This automated parameter transformation enables precise measurement of the device's ability to discern modulated speech without requiring complex manual test design for each voice variation.
Solution Approach 2:
The patent applies local quality by selectively modifying specific acoustic parameters of utterances while keeping others constant. This allows targeted testing of particular discrimination capabilities (e.g., pitch variation detection, timbre invariance) independently, enabling precise measurement of specific voice modulation discrimination abilities without overwhelming test complexity.
3Reliability
If the voice assistant device is tested in noisy environments, then real-world performance is better evaluated, but it becomes more challenging to determine accurate meaning of voice commands
Solution Approach 1:
The patent introduces an intermediary processing layer that separates noise addition from meaning evaluation. The system first generates clean modified utterances with controlled acoustic characteristics, then evaluates device responses to these predetermined phrases. This intermediary approach allows noisy environment testing while maintaining measurement precision by using controlled test inputs with known meanings, avoiding the ambiguity of uncontrolled real-world noise scenarios.
Data Source
AI summary
Computer-implemented methods, computer program products, and computer systems for testing a voice assistant device may include one or more processors configured for receiving test data from a database, wherein the test data may include a first set of coding parameters and a first user utterance having an expected device response. Further, the one or more processors may be configured for generating a first modified user utterance by applying the first set of coding parameters to the first user utterance, wherein the first modified user utterance is acoustically different than the first user utterance. The one or more processors may be configured for audibly presenting the first modified user utterance to a voice assistant device, receiving a first device response from the voice assistant device, and determining whether the first voice assistant response is substantially similar to the expected device response.