Voice code verification algorithm using competing models for user entrance authentication

In this paper, we propose a voice code verification method for an intelligent surveillance guard robot, wherein a robot prompts for a code (i.e. word or phrase) for verification. In the application scenario, the voice code can be changed every day for security reasoning and the targeting domain is unlimited. Thus, the voice code verification system not only requires the text-prompted and speaker independent verification, but also it should not require an extra trained model as an alternative hypothesis for log-likelihood ratio test because of memory limitation. To resolve these issues, we propose to exploit the subword based anti-models for log-likelihood normalization through reusing an acoustic model and competing with voice code model. The anti-model is automatically produced by using the statistical distance of phonemes against a voice code. In addition, a harmonics-based spectral subtraction algorithm is applied for a noisy robust system on an outdoor environment. The performance evaluation is done by using a common Korean database, PBW452DB, which consists of 63,280 utterances of 452 isolated words recorded in silent environment.