Voice-based authentication

A system comprising:

• one or more processors;
• and
• a computer-readable medium coupled to the one or more processors and comprising instructions thereon that, when executed by the one or more processors, cause the one or more processors to:
• obtain an audio clip having an utterance having an activation trigger portion and a command portion;
• select an activation trigger vector space model from multiple available activation trigger vector space model based on the activation trigger portion;
• create an activation trigger vector representation using the activation trigger vector space model;
• determine an activation trigger vector distance between the activation trigger vector representation of the activation trigger portion and a registered activation trigger vector;
• determine a command vector distance between a command vector representation of the command portion and a registered command vector;
• and
• responsive to the activation trigger vector distance satisfying an activation trigger distance threshold and the command vector distance satisfying a command vector distance threshold, providing service or access associated with a registered user.

A method comprising:

• receiving, from a device, an audio clip having an utterance;
• segmenting the utterance into an activation trigger portion and a command portion;
• creating an activation trigger vector representation of the activation trigger portion using a first vector space model;
• creating an command vector representation of the command portion using a second vector space model;
• provide service or access associated with a registered user based on the activation trigger vector representation and the command vector representation.