In a world dominated by smartphones, cloud computing, and online accounts, security of personal and corporate data is a critical concern. Voiceprint Vault provides a voice authentication service that can be used in a multitude of applications to secure sensitive data. Voiceprint Vault includes the following high-level features: - Cloud-based voice authentication using trusted signal processing algorithms - Multifactor authentication with use of optional password - Cross-platform compatibility using secure web requests to authenticate - Built-in storage and synching of private user data - Java library to facilitate integration with Android applications The Voiceprint Vault service allows users of an application to create an account, provide a voice sample, and then access the account with a simple spoken phrase. When users access their account, their voice sample is analyzed and compared to their training recordings. This system can be tailored to the needs of a particular user with per-user security options. It provides the convenience of voice access, but also allows for a password to be used for increased security. The Voiceprint Vault service is designed to allow application developers to integrate an existing, tested authentication system into their app rather than creating their own authentication system. The Voiceprint Vault server provides application specific repositories that developers can create to hold all user data, cryptographic information, and voice samples. The user data stored on the Voiceprint Vault server provides built-in synchronization across all connected devices. A reference implementation is provided that demonstrates the use of Voiceprint Vault authentication. The reference implementation is an Android app that uses the voice authentication service to protect access to personal notes, tasks, and dates that are synched across devices. Detailed instructions for integrating Voiceprint Vault into an existing application are also provided with the reference implementation. The accuracy of voiceprint authentication was investigated and optimized for a set of sample users and recordings. The security features and dangers of such a system are described along with recommendations for safe use. The optimal parameters to be used in the voice authentication algorithms are also presented in this report. / text

Voiceprints, like fingerprints, are a biometric. Where fingerprints record a person's unique pattern on their finger, voiceprints record what a person's voice "sounds like", abstracting away from what the person said. They have been used in speaker recognition, including verification and identification. In other words, they have been used to ask "is this speaker who they say they are?" or "who is this speaker?", respectively. However, people age, and so do their voices. Do voiceprints age, too? That is, can a person's voice change enough that after a while, the original voiceprint can no longer be used to identify them? In this thesis, I use Swedish audio recordings from Riksdagen's (the Swedish parliament) debate speeches to test this idea. Depending on the answer, this influences how well we can search the database for previously unmarked speeches. I find that speaker verification performance decreases as the age-gap between voiceprints increases, and that it decreases more strongly after roughly five years. Additionally, I grouped the speakers into age groups spanning five years, and found that speaker verification has the highest performance for those for whom the initial voiceprint was recorded from 29-33 years of age. Additionally, longer input speech provides higher quality voiceprints, with performance improvements stagnating when voiceprints become longer than 30 seconds. Finally, voiceprints for men age more strongly than those for women after roughly 5 years. I also investigated how emotions are encoded in voiceprints, since this could potentially impede in speaker recognition. I found that it is possible to train a classifier to recognise emotions from voiceprints, and that this classifier does better when recognising emotions from known speakers. That is, emotions are encoded more characteristically per person as opposed to per emotion itself. As such, they are unlikely to interfere with speaker recognition.