Engine for human language comprehension of intent and command execution

• US 9,961,442 B2
• Filed: 09/10/2015
• Issued: 05/01/2018
• Est. Priority Date: 11/21/2011
• Status: Active Grant

• Current Assignee: Zero Iabs, Inc.
• Original Assignee: Zero Iabs, Inc.
• Inventors: Pradhan, Rajesh, Pradhan, Amit
• Primary Examiner(s): VO, HUYEN X
• Application Number:
  US14/850,747
• Publication Number:
  US 20160005401A1
• Time in Patent Office: 964 Days
• Field of Search: 704233, 704200-231, 704272, 704270, 704500-504
• CPC Class Codes: G10L 15/18 using natural language modelling
• CPC Class Codes: G10L 2021/02166 Microphone arrays; Beamforming
• CPC Class Codes: G10L 21/0208 Noise filtering
• CPC Class Codes: H04R 3/005 for combining the signals of two or more microphones specially adapted for hearing aids H04R25/407

Abstract

The invention provides a computer system for interacting with a user. A set of concepts initially forms a target set of concepts. An input module receives a language input from the user. An analysis system executes a plurality of narrowing cycles until a concept packet having at least one concept has been identified. Each narrowing cycle includes identifying at least one portion of the language and determining a subset of concepts from the target set of concepts to form a new target subset. An action item identifier identifies an action item from the action items based on the concept packet. An action executer that executes an action based on the action item that has been identified.

Claims

1. A computer system for interacting with a user, comprising:

a processor;

a storage medium connected to the processor;

at least two microphones detecting sound to obtain a stereo input, the sound including speech from the user and background noise;

a set of instructions on the storage medium and executable by the processor, the set of instructions including;

a fast switcher;

determining whether sound can be detected through the two microphones simultaneously;

if the sound can be detected through the two microphones simultaneously then detecting the sound through the two microphones simultaneously; and

if the sound cannot be detected through the two microphones simultaneously then switching detection of the sound between the two microphones;

a triangulator analyzing the stereo input to isolate the speech from the sound using a triangulated position of the user relative to the two microphones; and

a dispatcher submitting the speech for analysis.

2. The computer system of claim 1, wherein the triangulator includes a convoluted triangulator that:

measures frequency of the sound in the time domain;

uses a fast Fourier transform to obtain sound magnitude in the frequency domain;

calculates an instantaneous magnitude across the frequency domain;

stores the instantaneous magnitude in a respective bin of a plurality of bins; and

groups the bins in groups having similar magnitude to isolate sources of the sound.

3. The computer system of claim 2, wherein the triangulator includes an instantaneous triangulator that:

executes a fast independent component analysis (FastICA) for each group.

4. The computer system of claim 1, further comprising:

at least one data store;

a set of concepts in the data store, the set of concepts initially forming a target set of concepts;

a plurality of action items in the data store;

an analysis system that executes a plurality of narrowing cycles until a concept packet having at least one concept has been identified, wherein each narrowing cycle includes;

i) identifying at least one portion of the language; and

ii) based on the portion of the language identifying a subset of concepts from the target set of concepts, the subset of concepts forming the target set of concepts to the exclusion of concepts previously forming part of the target set of concepts that are not based on the portion of the language;

an action item identifier that identifies an action item from the action items based on the concept packet; and

an action executer that executes an action based on the action item that has been identified.

5. A computer-based method of interacting with a user, comprising:

determining, with a processor, whether sound can be detected through at least two microphones simultaneously;

detecting, with the processor, sound using the two microphones to obtain a stereo input, the sound including speech from the user and background noise, wherein;

if the sound can be detected through the two microphones simultaneously then the detection of the sound is through the two microphones simultaneously; and

if the sound cannot be detected through the two microphones simultaneously the detection of the sound is by switching detection of the sound between the two microphones;

analyzing, with the processor, the stereo input to isolate the speech from the sound using a triangulated position of the user relative to the two microphones; and

dispatching with the processor, the speech for analysis.

6. A non-transitory computer-readable medium having storage thereon and a set of instructions executable by a processor of a computer to perform a computer-based method of interacting with a user, comprising:

determining, with a processor, whether sound can be detected through at least two microphones simultaneously;

detecting, with the processor, sound using the two microphones to obtain a stereo input, the sound including speech from the user and background noise, wherein;

if the sound can be detected through the two microphones simultaneously then the detection of the sound is through the two microphones simultaneously; and

if the sound cannot be detected through the two microphones simultaneously the detection of the sound is by switching detection of the sound between the two microphones;

analyzing, with the processor, the stereo input to isolate the speech from the sound using a triangulated position of the user relative to the two microphones; and

dispatching with the processor, the speech for analysis.