Emanuele Dalmasso

Emanuele Dalmasso
research category image

Main area of research
Biometrics

Emanuele Dalmasso is a Research Engineer on the Voice Biometrics team. His research interest is in the area of biometrics.

Prior to his career at Nuance, Emanuele co-founded BizCore Engineering, founded Dalmax.net, and worked with Ipse Docet and Loquendo. He also held a post-doctorate research position in collaboration with Politecnico di Torino, Italy, Loquendo and the I.S.I. Foundation. Emanuele has spent 10 years working in the voice biometric and speech analysis field, developing algorithms for Final State Network ASR, language identification and speaker segmentation and recognition. Emanuele also worked on developing highly optimized BOM software for the automotive industry, and he has more than three years of experience in mobile development for Android with some published apps. Emanuele is skilled in code profiling and optimization, research, algorithm design and implementation, GUI creation, and speech signal processing. His research interests include feature compensation, machine learning, and data analysis. Emanuele has published more than 10 papers in conferences and journals. He attended Politecnico di Torino, Italy, where he received an M.S. in computer engineering and a Ph.D. in computer and system engineering.

Selected articles

Loquendo-Politecnico Di Torino system for the 2009 NIST language recognition evaluation

This paper describes the system submitted by Loquendo and Politecnico di Torino (LPT) for the 2009 NIST Language Recognition Evaluation. The system is a combination

Read more

Compensation of nuisance factors for speaker and language recognition

The variability of the channel and environment is one of the most important factors affecting the performance of text-independent speaker verification systems. The best techniques

Read more

Stream-based speaker segmentation using speaker factors and eigenvoices

This paper presents a stream-based approach for unsupervised multi-speaker conversational speech segmentation. The main idea of this work is to exploit prior knowledge about the

Read more