Wake-Word-Free Utterance Processing With Intent Correlation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Voice assistants require a wakeup word for processing user instructions, leading to inconvenience and difficulty in distinguishing user utterances intended for the assistant from daily dialogues.
Innovation Solution
An electronic device processes utterances without a wakeup word by extracting domains and intents from initial and subsequent user inputs, determining correlations, and activating the microphone based on context and user history to process subsequent utterances.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If an always-on microphone is used to process utterances without a wakeup word, then user convenience is improved, but the ability to distinguish relevant utterances from daily dialog deteriorates
Solution Approach 1:
The system performs preliminary processing by extracting domains and intents from utterances before full activation. The microphone remains in a low-power listening state, performing preliminary analysis of incoming speech to determine if it warrants full processing, thus maintaining convenience while improving discrimination accuracy.
Solution Approach 2:
The system dynamically adjusts its processing state based on the characteristics of the detected utterance. It transitions between different operational modes (listening state vs. full processing state) depending on whether the utterance matches relevant domains and intents, allowing adaptive response to different speech patterns.
2Speed
If the microphone is always active to process any user utterance, then response speed is improved, but energy consumption increases
Solution Approach 1:
The system employs periodic action by switching between active listening periods and full processing periods. The microphone operates in a low-power periodic listening mode, activating full processing only when relevant domains or intents are detected, thereby reducing overall energy consumption while maintaining fast response capability.
Solution Approach 2:
The system performs partial processing by extracting only domain and intent information from utterances in the listening state, rather than performing complete speech recognition and processing for all utterances. This partial action reduces energy consumption while maintaining the ability to quickly respond to relevant inputs.
3Measurement precision
If domain and intent extraction is performed for every utterance, then processing accuracy is improved, but computational complexity increases
Solution Approach 1:
The system applies local quality by performing domain and intent extraction selectively rather than uniformly for all utterances. Full extraction and processing is applied locally only to utterances that show characteristics of relevance based on preliminary analysis, reducing overall computational complexity while maintaining high accuracy for relevant inputs.
Data Source
AI summary
An electronic device and an utterance processing method of the electronic device are provided. The electronic device includes a microphone that receives a first and a second utterance of a user, the second utterance being generated at a later point in time than the first utterance, a memory storing instructions, and at least one processor. The at least one processor may be configured to execute the instructions to extract a first domain corresponding to the first utterance and a first intent corresponding to the first utterance, determine a correlation between the first utterance and the second utterance, and determine whether to process the second utterance based on the correlation.