Back to results
US 10,854,192 B1Granted
Domain specific endpointing
AssigneeAmazon Technologies, Inc.
InventorsRoland Maas, Ariya Rastrow, Rohit Prasad
Life of this patent
Estimating the term…
Abstract
An automatic speech recognition (ASR) system detects an endpoint of an utterance based on a domain of the utterance. The ASR system processes a first portion of the utterance to determine the domain and then determines an endpoint of the remainder of the utterance depending on the domain.
Claims18
A computer-implemented method for determining an utterance endpoint during automatic speech recognition (ASR) processing, the computer-implemented method comprising:
- receiving audio data comprising a first portion and a second portion, the audio data corresponding to an utterance;
- performing ASR processing on the first portion to determine a first interpretation;
- determining that the first interpretation includes a command;
- determining, using the first interpretation, a domain of the command;
- identifying a threshold quantity of consecutive non-speech audio frames corresponding to an endpoint corresponding to the domain;
- performing ASR processing on the second portion to determine a second interpretation;
- determining that the second portion includes a first quantity of consecutive non-speech audio frames;
- determining an expected pause duration based at least in part on the second interpretation;
- determining that the first quantity of consecutive non-speech audio frames exceeds the threshold quantity of consecutive non-speech audio frames;
- determining that the first quantity of consecutive non-speech audio frames exceeds the expected pause duration;
- and
- determining, based at least in part on the domain and the expected pause duration, an utterance endpoint in the audio data.
9 more elements
A computer-implemented method comprising:
- receiving first audio data representing speech;
- performing automatic speech recognition (ASR) processing on the first audio data to determine the speech represents a command;
- determining a domain corresponding to the command;
- determining, based at least in part on the domain, a threshold non-speech duration of audio data;
- and
- determining, based at least in part on the threshold non-speech duration, a speech endpoint in the first audio data.
3 more elements
A computing system comprising:
- at least one processor;
- and
- a memory including instructions that, when executed by the at least one processor, cause the computing system to:
- receive first audio data representing speech;
- perform automatic speech recognition (ASR) processing on the first audio data to determine the speech represents a command;
- determine a domain corresponding to the command;
- determine, based at least in part on the domain, a threshold non-speech duration of audio data;
- and
- determine, based at least in part on the threshold non-speech duration, a speech endpoint in the first audio data.
6 more elements
Dependencies are inferred from the claim text.