Text-prompted speaker recognition with polynomial classifiers
نویسندگان
چکیده
A novel system for text-prompted speaker recognition is presented. The system first segments the speech by Viterbi alignment with speaker independent models. It then applies a polynomial classifier to each subword for recognition. This methodology has several interesting aspects. First, the system has excellent computational scalability for identification. Second, the discriminative training method incorporates the background normalization into the enrollment process. Third, training can be performed with one-pass through the enrollment data. Experiments show that the new system is competitive with current HMM based approaches.
منابع مشابه
"text-prompted" without Text: a Language-independent Voice-prompted Speaker Recognition System
A new paradigm of voice prompted speaker recognition is presented. The vocal prompts that the speaker is asked to repeat are used by the speaker recognition system for segmenting the data and for normalizing the verification results. Using the vocal prompts themselves instead of the matching text makes the overall system more flexible and truly language independent. A technology demonstration s...
متن کاملSpeaker Recognition Using Auditory Features and Polynomial Classifier
This paper presents a speaker recognition method which makes use of auditory features and polynomial classifier for speaker recognition. Auditory features based on an auditory periphery model extract significant speaker characteristics. Polynomial classifier has been used to accomplish speaker recognition task. Polynomial classifier has several advantages over the conventional classifiers such ...
متن کاملSpeaker identification using a polynomial-based classifier
A new set of techniques for using polynomial-based classifiers for speaker identification is examined. This set of techniques makes application of polynomial classifiers practical for speaker identification by enabling discriminative training for large data sets. The training technique is shown to be invariant to fixed liftering and affine transforms of the feature space. Efficient methods for ...
متن کاملSpeaker verification with ensemble classifiers based on linear speech transforms
For most classifier architectures realistic training schemes only allow classifiers corresponding to local optima of the training criteria to be constructed. One way of dealing with this problem is to work with classifier ensembles: multiple classifiers are trained for the same classification problem and combined into one “super” classifier. The problem addressed in this paper is text prompted ...
متن کاملMethods of Combining Multiple Classifiers with Different Features and Their Applications to Text-Independent Speaker Identification
In practical applications of pattern recognition, there are often different features extracted from raw data which needs recognizing. Methods of combining multiple classifiers with different features are viewed as a general problem in various application areas of pattern recognition. In this paper, a systematic investigation has been made and possible solutions are classified into three framewo...
متن کامل