Self-attention encoding and pooling for speaker recognition
| Autor: | Pooyan Safari, Javier Hernando, Miquel India |
|---|---|
| Přispěvatelé: | Universitat Politècnica de Catalunya. Doctorat en Teoria del Senyal i Comunicacions, Universitat Politècnica de Catalunya. Departament de Teoria del Senyal i Comunicacions, Universitat Politècnica de Catalunya. VEU - Grup de Tractament de la Parla |
| Rok vydání: | 2020 |
| Předmět: |
FOS: Computer and information sciences
Computer Science - Machine Learning 0209 industrial biotechnology Sound (cs.SD) Computer science Self-attention encoding Speech recognition Pooling 02 engineering and technology Computer Science - Sound Machine Learning (cs.LG) Reduction (complexity) 020901 industrial engineering & automation Discriminative model Natural language processing (Computer science) Audio and Speech Processing (eess.AS) Encoding (memory) 0202 electrical engineering electronic engineering information engineering FOS: Electrical engineering electronic engineering information engineering Tractament del llenguatge natural (Informàtica) Representation (mathematics) Transformer (machine learning model) Speaker recognition Speaker embedding Self-attention pool-ing Speaker verification Enginyeria de la telecomunicació::Processament del senyal::Processament de la parla i del senyal acústic [Àrees temàtiques de la UPC] 020201 artificial intelligence & image processing Mobile device Electrical Engineering and Systems Science - Audio and Speech Processing |
| Zdroj: | UPCommons. Portal del coneixement obert de la UPC Universitat Politècnica de Catalunya (UPC) INTERSPEECH |
| DOI: | 10.48550/arxiv.2008.01077 |
| Popis: | The computing power of mobile devices limits the end-user applications in terms of storage size, processing, memory and energy consumption. These limitations motivate researchers for the design of more efficient deep models. On the other hand, self-attention networks based on Transformer architecture have attracted remarkable interests due to their high parallelization capabilities and strong performance on a variety of Natural Language Processing (NLP) applications. Inspired by the Transformer, we propose a tandem Self-Attention Encoding and Pooling (SAEP) mechanism to obtain a discriminative speaker embedding given non-fixed length speech utterances. SAEP is a stack of identical blocks solely relied on self-attention and position-wise feed-forward networks to create vector representation of speakers. This approach encodes short-term speaker spectral features into speaker embeddings to be used in text-independent speaker verification. We have evaluated this approach on both VoxCeleb1 & 2 datasets. The proposed architecture is able to outperform the baseline x-vector, and shows competitive performance to some other benchmarks based on convolutions, with a significant reduction in model size. It employs 94%, 95%, and 73% less parameters compared to ResNet-34, ResNet-50, and x-vector, respectively. This indicates that the proposed fully attention based architecture is more efficient in extracting time-invariant features from speaker utterances. This work was supported in part by the Spanish Project DeepVoice (TEC2015-69266-P). |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|