Accurately Identify Sounds, Speakers, & Scenes

Sensory’s Sound Identification, SoundID, represents the next step in the evolution of speech recognition. SoundID enables developers and OEMs to create devices that go beyond sensing the human voice and recognize specific sounds in an environment. The four core SoundID capabilities; Sound Event Detection, Sound Scene Tracking, Alarm Detection, and Enrolled Sound Detection combined offer a multitude of possible applications from smart home alert systems to industrial equipment safety monitoring.  SoundID, a feature of the powerful TrulySecure™ embedded AI platform, leverages a novel two-stage detection approach to provide the perfect balance between on-device computation and overall recognition performance.

 

 

Image Alt Text

EXPANSIVE SOUNDID LIBRARY

SoundID’s current library includes 16 sample sounds with various combinations (home, health, safety, etc.) Sound Scene Tracking, enables listening continuously to audio and reporting relative match to broad sound categories or “scenes”. Custom scenes are available by request.   End users may add their own sounds via Enrolled Sound Detection, which can help devices recognize unique household sounds that are discrete (e.g doorbells), or continuous (e.g. running water).

 

Image Alt Text

Key Features

Available on device or via a client/cloud hybrid using an always-on low power technology for the ultimate in privacy without sacrificing accuracy.

Listens for predetermined sounds like glass break, gunshots, dog barks, snoring, alarms, and more!
Listens for predetermined sounds like glass break, gunshots, dog barks, snoring, alarms, and more!
Ability to build industrial sound detection for specific product issues or malfunctions
Ability to build industrial sound detection for specific product issues or malfunctions
Custom end user sounds can be sampled or adapted on device
Custom end user sounds can be sampled or adapted on device
Multi-stage shallow and deep learning for lowest power consumption and highest accuracy
Multi-stage shallow and deep learning for lowest power consumption and highest accuracy
Runs embedded on various OS systems (Linux, Android, Windows, etc.) or can be custom ported to low power DSPs and microcontrollers
Runs embedded on various OS systems (Linux, Android, Windows, etc.) or can be custom ported to low power DSPs and microcontrollers

TrulySecure Speaker Verification

Sensory’s TrulySecure Speaker Verification (TSSV) SDK provides a real-time, on-device, voice biometric API targeting a wide range of computing platforms. The SDK also supports wake-word detection and includes SoundID.

Request the TrulySecure Speaker Verification product brief to learn more.

Image Alt Text)

How it works

Through a mix of deep and shallow learning, Sensory AI technology can now recognize and identify specific predetermined sounds, custom sounds, and “scenes” to better understand a user’s environment. The technology is available on device or via a client/cloud hybrid using an always-on low power technology for the ultimate in privacy without sacrificing accuracy.

All sound analyses are done on-device, so homeowners don’t have to worry about recordings being shared in any way. TrulySecure Sound ID uses multiple approaches to identifying sound in a way that’s more accurate than any single approach could be on its own. 

Download the SoundID product brief to learn more.

Image Alt Text)

Stay Connected

Get the latest speech recognition and artificial intelligence resources and industry updates.