Comparison of text - independent speaker recognition methodson telephone speech with
نویسنده
چکیده
We compare speaker recognition performance of Vector Quantization (VQ), Gaussian Mixture Modeling (GMM) and the Arithmetic Harmonic Sphericity measure (AHS) in adverse telephone speech conditions. The aim is to address the question: how do multimodal VQ and GMM typically compare to the simpler unimodal AHS for matched and mismatched training and testing environments. We study identiication (closed set) and veriication errors on a new multi-environment database. We consider LPC and PLP features as well as their RASTA derivatives. We conclude that RASTA processing can remove redundancies from the features. We aarm that even when we use channel and noise compensation schemes speaker recognition errors remain high when there is acoustic mismatch.
منابع مشابه
Text-independent Speaker Recognition by Trajectory Space Comparison
We present the principle of trajectory space comparison for text-independent speaker recognition and some solutions to the space comparison problem based on vector quantization. The comparison of recognition rate of diierent solutions is reported. Experimental system achieved 99.5% text-independent speaker recognition rate for 23 speakers, using 5 phrases for training and 5 for test. A speaker-...
متن کاملText-independent Forensic Speaker Identification Using Telephone Speech
On several occasions we have been asked by police officials whether it is feasible for investigative purposes to automatically identify a speaker using speech from intercepted telephone conversations. Since forensic phoneticians are typically not involved in speaker identification but in speaker verification we tend to reformulate this question in : Is it feasible for investigative purposes to ...
متن کاملSpeaker Adaptation in Continuous Speech Recognition Using MLLR-Based MAP Estimation
A variety of methods are used for speaker adaptation in speech recognition. In some techniques, such as MAP estimation, only the models with available training data are updated. Hence, large amounts of training data are required in order to have significant recognition improvements. In some others, such as MLLR, where several general transformations are applied to model clusters, the results ar...
متن کاملComparison of text-independent speaker recognition methods on telephone speech with acoustic mismatch
We compare speaker recognition performance of Vector Quantization (VQ), Gaussian Mixture Modeling (GMM) and the Arithmetic Harmonic Sphericity measure (AHS) in adverse telephone speech conditions. The aim is to address the question: how do multimodal VQ and GMM typically compare to the simpler unimodal AHS for matched and mismatched training and testing environments. We study identi cation (clo...
متن کامل