Automated Assistant Hot Command Enrollment via Preliminary Speech Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automated assistants require explicit invocation and transition into a full listening state to respond to user commands, which can be cumbersome and inefficient, especially when users need to utter long tail hot words.
Innovation Solution
The implementation of techniques to detect and enroll new 'hot commands' that allow automated assistants to perform responsive actions without explicit invocation, by performing speech recognition processing on utterances captured at times other than immediately after invocation, and analyzing textual snippets to determine if they should trigger responsive actions or be enrolled as hot commands.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Use of energy by moving object
If the automated assistant operates in a limited listening state with a finite set of hot words, then power consumption and computing resources are conserved, but user convenience deteriorates due to the need to utter long tail hot words for invocation
Solution Approach 1:
The system performs preliminary speech recognition processing on captured audio data to generate textual snippets before explicit invocation. This preliminary action enables the system to identify potential hot commands in advance, so when users utter these commands, the assistant can respond immediately without requiring full invocation, thus improving convenience while maintaining energy efficiency through selective processing.
Solution Approach 2:
The system applies partial speech recognition processing by generating textual snippets from audio data without always transitioning to the full listening state. This partial action allows the system to identify common command patterns and enroll them as hot commands, reducing the need for complete invocation sequences and improving ease of operation while conserving resources.
2Productivity
If the automated assistant performs speech recognition processing only after explicit invocation, then computing resources are conserved, but responsiveness deteriorates for frequently used commands
Solution Approach 1:
The system performs preliminary speech recognition processing on audio data to generate textual snippets before explicit invocation. This preliminary action identifies frequently used commands, which are then enrolled as hot commands. When these commands are detected, the system responds immediately without requiring full invocation and transition to listening state, thus improving responsiveness while managing processing complexity through selective enrollment.
3Adaptability or versatility
If the automated assistant enrolls new hot commands from user utterances, then adaptability improves, but system complexity increases due to continuous learning and enrollment processes
Solution Approach 1:
The system applies partial processing by generating textual snippets from audio data and analyzing them for potential hot command enrollment without implementing complete continuous learning. The enrollment process focuses on identifying frequently occurring commands that meet specific criteria, rather than continuously adapting to all possible user utterances. This approach improves adaptability for commonly used commands while limiting system complexity by avoiding comprehensive continuous learning mechanisms.
4Productivity
If the automated assistant transitions to full listening state for every user utterance, then responsiveness improves, but power consumption increases
Solution Approach 1:
The system performs preliminary speech recognition processing on captured audio data to generate textual snippets before transitioning to full listening state. This preliminary action identifies potential hot commands, allowing the system to respond immediately to frequently used commands without full invocation, thus maintaining responsiveness while reducing power consumption by avoiding unnecessary state transitions.
Solution Approach 2:
The system applies partial speech recognition processing by generating textual snippets from audio data and identifying potential hot commands without always transitioning to the full listening state. This partial action enables the system to respond to enrolled hot commands efficiently while conserving power by limiting full processing to only when necessary.
Data Source
AI summary
Techniques are described herein for detecting and/or enrolling (or commissioning) new “hot commands” that are useable to cause an automated assistant to perform responsive action(s) without having to be first explicitly invoked. In various implementations, an automated assistant may be transitioned from a limited listening state into a full speech recognition state in response to a trigger event. While in the full speech recognition state, the automated assistant may receive and perform speech recognition processing on a spoken command from a user to generate a textual command. The textual command may be determined to satisfy a frequency threshold in a corpus of textual commands. Consequently, data indicative of the textual command may be enrolled as a hot command. Subsequent utterance of another textual command that is semantically consistent with the textual command may trigger performance of a responsive action by the automated assistant, without requiring explicit invocation.