Malaspina Labs announces industry’s first biometric speaker verification solution to execute entirely on ultra-low power devices

September 15, 2015 — Malaspina Labs announces the release of the industry’s first biometric speaker verification solution to execute entirely on an ultra-low power target device without the need for pairing the device with a mobile phone, computer, or the Internet.
The solution will be demonstrated at the ARC Processor Summit, September 15, 2015.

Malaspina’s VoiceBoost™ suite of speech processing solutions includes noise reduction, voice activation, biometric speaker verification, and embedded speech recognition products. The VoiceBoost™ Biometric Speaker Verification solution allows devices to implement voice authentication, so that a device will respond only when a user’s passphrase is spoken by the user’s own voice. This allows manufacturers to add device locking, parental controls and multi-user capabilities to smart consumer and IoT devices.

Speech is an appealing interface option for consumer devices that do not have a keyboard or touch screen by which to access device functions. For device security and user identification purposes, Malaspina Labs and others offer biometric speaker verification solutions that can authenticate users by recognizing the unique sound of each user’s voice. Before a user’s voice can be recognized by these solutions, each user must “enroll” his or her voice and chosen passphrase. However, this voice enrollment step has traditionally required many megabytes of memory and hundreds of MIPS of processing power. Although this is not a hindrance on smartphones, the computing and memory requirements of the voice enrollment step has prevented biometric speaker verification solutions from executing on lower power consumer devices unless that device is connected with a mobile phone or computer during the voice enrollment process. Malaspina’s new on-device speaker enrollment feature marks an industry milestone by eliminating the need for consumer devices to be connected to an external device to complete the voice enrollment process.

Malaspina’s new on-device speaker enrollment feature requires less than 90KB of memory and 50 Mcycles for real-time voice enrollment, making it an ideal solution for low power consumer devices lacking WiFi or Bluetooth capabilities or where pairing the device with an external system to enroll users’ voices is not desired. On-device enrollment and host-based enrollment options are both standard features of Malaspina Labs’ VoiceBoost™ Biometric Speaker Verification solution optimized for Synopsys’ DesignWare® ARC® EM DSP processors.

Once one or more users have enrolled their voices, Malaspina’s Biometric Speaker Verification solution can execute in an ultra-low power “always listening” mode, recognizing individual users by their voices even with background noise or with users speaking at arm’s length and beyond. This capability enables new functionality for low-power consumer devices such as parental controls for home automation devices, different commands for different users, and collecting & keeping multiple users’ information separate on the device.

Malaspina has partnered with Synopsys to optimize Malaspina’s VoiceBoost suite of speech solutions for Synopsys’ ARC EM DSP family of processors. ARC processors are specifically designed for embedded devices requiring a high degree of power-performance efficiency, such as always-on, voice activated and sensor processing applications. The Biometric Speaker Verification solution consumes just 2.25 μW of power on average in typical conditions on ARC EM DSPs.

“Voice biometrics for authentication provide a higher level of security than traditional methods and are a key feature for many emerging IoT devices in the market“ said Matt Gutierrez, Director of Marketing for processor solutions at Synopsys. “By collaborating with Malaspina Labs to optimize their on-device voice enrollment technology for Synopsys’ ARC EM processors, we are enabling new classes of products on which users can enroll and use VoiceBoost Biometric Speaker Verification without the need to connect to a host".

Dean Neumann, CEO of Malaspina Labs said “Malaspina is pleased to offer the industry’s first completely on-device speaker verification solution for smart devices, and optimize it for the Synopsys ARC EM DSP family of ultra-low power embedded processors. This allows manufacturers to provide more secure device access than PIN codes and allows consumer devices to implement embedded speech recognition with user-specific functionality and control, without the need to pair the device with any external host for voice enrollment. Imagine a garage door opener that can authenticate a user by voice, or an MP3 player that knows who’s playlist to play, or a heart rate monitor that knows who’s target heart rate and training profile to use ... This is a game changer."

About Malaspina Labs

Malaspina Labs develops speech interface and enhancement software for ultra-low power devices which improves person-to-person and person-to-device voice communication in high noise environments. Malaspina Labs' VoiceBoost™ suite of solutions provides speech isolation and voice interface technology that is effective at handset, speakerphone and arm's length distances and beyond. VoiceBoost solutions operate deterministically at bone-conduction, telecom, and voice interface latencies within the constraints of milliwatt-class processors designed for mobile, embedded, medical and consumer devices.

For more information about Malaspina Labs, visit