Smart Home Noise Signature Database for Voice Command Recognition

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Smart home devices face challenges in effectively filtering out background noise from voice commands due to the short duration and abrupt nature of these commands, as traditional noise cancellation algorithms require a convergence interval to distinguish between background and foreground sounds, often missing or misinterpreting voice commands.

Innovation Solution

Implementing a system that detects the start of a voice command and applies a pre-selected background noise signature from a database of collected noise profiles, allowing immediate noise cancellation without waiting for convergence, using environmental inputs like time of day and appliance status to select the appropriate noise signature.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If traditional noise cancellation algorithms are used, then background noise can be filtered out over time, but voice commands with short duration are missed or misinterpreted due to the convergence interval requirement

Engineering Contradiction:

Improvevoice command recognition accuracyVSAvoidconvergence interval duration

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The system performs preliminary action by pre-selecting and storing multiple background noise signatures corresponding to different environmental conditions (e.g., kitchen, living room, time of day) before voice commands are issued. When a voice command is detected, the appropriate pre-selected signature is immediately applied without requiring a convergence interval, thus resolving the contradiction between reliability and time loss.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system implements dynamics by making the background noise signature adaptable and changeable based on environmental inputs. Multiple noise signatures are stored and dynamically selected based on current environmental conditions (appliance status, time of day, location), allowing the system to adapt to varying acoustic environments without requiring re-convergence, thereby improving voice command recognition while minimizing time loss.

Inventive Principle:

Principle #15Dynamics

2Measurement precision

If noise cancellation converges slowly to distinguish background from foreground sounds, then accurate noise filtering is achieved, but short voice commands are missed or misinterpreted

Engineering Contradiction:

Improvenoise discrimination accuracyVSAvoidconvergence time

Core Design Contradiction:

Measurement precisionVSDuration of action of moving object

Solution Approach 1:

The system performs preliminary action by pre-characterizing background noise signatures under various environmental conditions and storing them in advance. When a voice command is detected, the matching pre-characterized signature is immediately applied, eliminating the need for slow convergence while maintaining accurate noise discrimination. This resolves the contradiction between measurement precision and duration of action.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system prepares beforehand by storing multiple pre-selected noise signatures that cushion against the variability of different environmental acoustic conditions. This prior preparation allows immediate application of appropriate noise profiles without requiring time-consuming convergence, thus maintaining measurement precision while reducing convergence time.

Inventive Principle:

Principle #11Beforehand cushioning (Prior cushioning)

3Reliability

If the system records and processes long-duration background noise for convergence, then accurate noise profiles are established, but short voice commands cannot be properly distinguished

Engineering Contradiction:

Improvenoise profile accuracyVSAvoidvoice command response speed

Core Design Contradiction:

ReliabilityVSProductivity

Solution Approach 1:

The system performs preliminary action by pre-recording and storing multiple background noise signatures during different environmental conditions (e.g., with various appliances on, at different times of day) before they are needed. This pre-characterization maintains reliable noise profile accuracy while enabling immediate application when voice commands are detected, thus resolving the contradiction between reliability and productivity.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system implements dynamics by making noise profile selection adaptive based on environmental inputs. Multiple noise signatures are dynamically selected and applied based on current conditions (appliance status, time of day, location), allowing the system to maintain accurate noise profiles while responding immediately to voice commands without requiring long-duration processing.

Inventive Principle:

Principle #15Dynamics

Data Source

PatentUS10178474B2Sound signature database for initialization of noise reduction in recordings

Publication Date: 2019.01.08 GOOGLE LLC

AI summary

A smart-home device may include a recording device configured to record sound during a first time interval and a memory device comprising a plurality of stored sound profiles. The smart-home device may also include a processing system configured to receive an environmental input, select a stored sound profile from the plurality of stored sound profiles based on the environmental input, and perform a noise-cancelation routine on the sound recorded during the first time interval. The stored sound profile may be used as an initial background noise profile for the noise-cancelation routine.