I4U Submission to NIST SRE 2018: Leveraging from a Decade of Shared Experiences

Abstract

The I4U consortium was established to facilitate a joint entry to NIST speaker recognition evaluations (SRE). The latest edition of such joint submission was in SRE 2018, in which the I4U submission was among the best-performing systems. SRE'18 also marks the 10-year anniversary of I4U consortium into NIST SRE series of evaluation. The primary objective of the current paper is to summarize the results and lessons learned based on the twelve sub-systems and their fusion submitted to SRE'18. It is also our intention to present a shared view on the advancements, progresses, and major paradigm shifts that we have witnessed as an SRE participant in the past decade from SRE'08 to SRE'18. In this regard, we have seen, among others, a paradigm shift from supervector representation to deep speaker embedding, and a switch of research challenge from channel compensation to domain adaptation.

Public Key Compression for Constrained Linear Signature Schemes

“…Still, if better code based signature schemes are developed, the construction will likely to be able to improve the quantity |sig|+|pk|. For example, even though the pqsigRM [22] proposal to the NIST PQC project does not have a completely unstructured matrix as public key, our construction can still reduce |sig| + |pk| by a factor 6 from 329 kB to 60 kB in this case (with α = 4, σ = 64). Unfortunately, comments on the NIST forum indicate that the pqsigRM proposal might not be secure [2].…”

Towards Robust Speaker Verification with Target Speaker Enhancement

“…Despite the success achieved, current SV systems still suffer evident performance degradation in the presence of severe background noise with interfering speakers [11,12,13,14,15]. Therefore, the SV system is at increased risk of failure when exposed to unconstrained conditions.…”

Detection of special scams by emotion analysis

“…However, this method may falsely detect normal conversations with similar content, so it needs to be combined with other analysis methods. In voiceprint analysis [3], a caller is identified by extracting the characteristics of the user's voice and comparing them with the characteristics registered in a database in advance. However, since this method analyzes the scammer's voice, the detection accuracy may decrease as the accuracy of voiceprint mimicry technology improves [4].…”