The IBM Speaker Recognition System: Recent Advances and Error Analysis

Abstract

We present the recent advances along with an error analysis of the IBM speaker recognition system for conversational speech. Some of the key advancements that contribute to our system include: a nearest-neighbor discriminant analysis (NDA) approach (as opposed to LDA) for intersession variability compensation in the i-vector space, the application of speaker and channel-adapted features derived from an automatic speech recognition (ASR) system for speaker recognition, and the use of a DNN acoustic model with a very large number of output units (∼10k senones) to compute the frame-level soft alignments required in the i-vector estimation process. We evaluate these techniques on the NIST 2010 SRE extended core conditions (C1-C9), as well as the 10sec-10sec condition. To our knowledge, results achieved by our system represent the best performances published to date on these conditions. For example, on the extended tel-tel condition (C5) the system achieves an EER of 0.59%. To garner further understanding of the remaining errors (on C5), we examine the recordings associated with the low scoring target trials, where various issues are identified for the problematic recordings/trials. Interestingly, it is observed that correcting the pathological recordings not only improves the scores for the target trials but also for the nontarget trials.

An Analysis of Transfer Learning for Domain Mismatched Text-independent Speaker Verification

“…In speaker recognition, i-Vector/PLDA systems achieve very impressive performances in the presence of noise and channel variabilities [1,2,3,4]. However, these systems often rely on a large collection of in-domain and well-annotated data, e.g., transcriptions for ASR DNN acoustic modeling and speaker labels for PLDA training [5,6].…”

“…UT-SCOPE-physical speech corpus is employed as the indomain data [19]. Instead of focusing on duration, noise or channel variabilities like the conventional NIST SREs [3,20], the corpus addresses more on the variability introduced by speakers themselves (i.e., intrinsic neutral/physical stressed mismatch). To create the mismatched conditions, neutral-read speech, stressed-read speech and stressed-spontaneous speech are collected from each speaker.…”