VoiceBoost™ Speech Command and Control

Low power embedded automatic speech recognition for DSPs, microcontrollers and SOCs allows high accuracy voice commands entirely on device

Mobile, wearable and automation devices continue to benefit the consumer by packing more functions into smaller, lower power, and less expensive products. However, smaller products lack keyboards and large touch screen displays for user input, resulting in cumbersome user interfaces for accessing the functions of the device. Smaller products also mean smaller batteries, resulting in a need for an alternate interface to control the functions of the device while operating at low power to maximize battery life.

Add Voice Commands to Wearable and Automation Devices

Malaspina Labs’ VoiceBoost™ Speech Command and Control solution allows manufacturers to add a speech interface to devices which executes in real-time entirely on the device’s DSP, microcontroller or SOC, requiring no connection to a cloud-based or host application.

The Speech Command and Control solution has ultra-small memory footprint and minimal power requirements and is designed specifically to address the needs of low-cost mobile, wearable and IoT devices. The solution recognizes any combination of discrete words, connected phrases and continuous speech, spoken by any user either near-field or from a distance. The Speech Command and Control solution is designed for robust performance in noisy environments and with speech from a distance or with off-axis speech, as is typical with wrist-worn devices. Market-leading recognition levels ensure robust performance, while OEM-selectable recognition confidence thresholds allow the application to request user confirmation whenever desired.

The Speech Command and Control solution can be used standalone or in combination with VoiceBoost “always listening” voice activation solutions. When used in combination with the speaker-independent voice activation solution, a device can remain in an ultra-low power state while continuously listening for a wakeup word or phrase from any user, after which the user may issue voice commands to control the device. This allows manufacturers to offer a complete speech command and control interface which works “out of the box” for any user while still providing always-listening operation with less than 1mA average power consumption.

When used in combination with the biometric speaker verification solution, a device can remain in an ultra-low power state while continuously listening for wakeup words spoken only by an enrolled user’s voice, after which the user may issue voice commands to control the device. Use of the Speech Command and Control solution in combination with the Biometric Speaker Verification solution allows manufacturers to support user-specific device functionality such as parental controls and device unlock, as well as adding multi-user capabilities such as personalized functions and information to mobile, wearable and IoT devices, while still consuming less than 1mA average power.