VoiceBoost™ Voice Activation and Verification

“Always listening” audio signal processing for low power audio processors allows hands-free device wakeup in noisy environments

Use of voice interfaces, “hands-free” modes and voice searches on mobile devices continues to grow. One requirement for a voice interface is that the device must first be in a state where it is actively listening for voice commands before a voice command can be spoken. When the phone is in its low-power "sleep" state, this can pose a problem. Currently, waking the phone from its standby state to its active state requires physical interactions such as a gesture or button press, followed by entering a PIN code or a password. Following wakeup, the device may need further interaction to be in the “listening” mode for voice input, such as an app launch. This need for a "hands-on" phycial interaction to wake the device before voice can be used is a major impediment to hands-free and eyes-free interaction.

To eliminate these physical interactions, an effective solution to wake the device from its sleep state with a "wakeup" voice keyword or keyphrase is needed. This allows the device to be "always listening" for voice wakeup commands, even in the standby state. An effective voice activation solution needs to execute continuously in a low-power state while effectively recognizing spoken wakeup commands from users, including users with accents and non-native language speakers, while being effective even at arm's length or in high noise environments.

Robust Hands-Free Wakeup Commands for Voice Activation

VoiceBoost Voice Activation solutions continuously listen for a keyword or keyphrase spoken by the user to activate the device and place it in a mode where its speech recognition application is ready to accept voice input. The solutions execute on low-power audio processors and can be active in an “always listening” mode even with the host processor in a standby state.

The VoiceBoost Speaker Independent Voice Activation solution allows an OEM to define a default keyphrase to wake up the device so that it works “out of the box” for any user. The solution offers keyword spotting with ultra-low power continuous listening capability for supported audio DSPs.

The VoiceBoost Speaker Dependent Voice Activation solution provides two-factor speaker verification using user-defined activation commands, and offers keyword spotting with ultra-low power continuous listening capability for supported audio DSPs. Users define their own unique keyphrases in their own voice, so that the device wakes up only upon hearing the user’s chosen password spoken by that user’s voice.

Both Speaker Independent and Speaker Dependent solutions offer market-leading key phrase detection rates in both low noise and high noise environments.

Speaker Verification: Malaspina Labs vs. market leading solution

Speaker Independent Voice Activation: Malaspina Labs vs. market leading solution