The Gaussian mixture based GMM-UBM approaches have shown good performance in speaker verification without using contextual information. This paper exploits the information provided in the arcs of a decoded syllable lattice for speaker verification. The forward algorithm is used to summarize this information in the syllable lattice instead of the best decoded string. The performance is evaluated on a Mandarin Chinese database. With two minutes of target speaker's enrollment data, the proposed algorithm shows 1.03% of equal-error rate for short input utterances with an average duration of two seconds. By combining with the GMM-UBM, the system shows a 0.74% of equal-error rate.
Related white papers
Methods for Measuring Perceptual Speech Quality
Describes four methods for measuring speech clarity and quality. Perceptual Speech Quality Measurement, Measuring Normalizing Blocks, Perceptual Speech Quality Measurement Plus, and Perceptual Analysis Measurement System are explained in detail....
Innovent Case Study: TradeHarbor
On August 3, 2004, TradeHarbor, a leading provider of voice authentication services, announced that Innovent, an entrepreneurial innovation unit within Nokia, had invested in the early-stage company and signed an...
TechNet Webcast: Planning and Deploying Voice Routes in Communications Server 2007 Using Route Helper (Level 300)
Enterprise Voice Route Helper is a Graphical User Interface (GUI) application that helps visualize, test, modify, archive, and share voice routing configuration information. Route Helper is an alternative to the...
Mainstreaming Speech-Enabled Web Applications
This white paper outlines the evolution of speech application development from these days static content to dynamically generated VoiceXML. As developers gain access to interoperable voice components within Web programming...
Healthcare Provider Increases Preventive Care, Boosts Efficiency With Speech Solution
Based in Monterey Park, California, Care1st Health Plan is a growing healthcare network of physicians and hospitals. The organization provides medical benefits to more than 200,000 low-income and disabled members....
Bathroom Retailer Supports Online Sales With New Phone Infrastructure
Taurus Bathrooms has operated a successful, family-run business for 25 years. The existing Taurus telephone system could not cope with the new demands of the internet business. They didn't physically...
TechNet Webcast: Speech Recognition and Grammar Generation in Unified Messaging for Exchange Server 2007 (Level 300)
A good unified messaging experience hinges on speech recognition. The presenter of this webcast talks about the speech recognition feature of Microsoft Exchange Server 2007 Unified Messaging. The focus is...


