Spoken Language Identification using ConvNets

9 Oct 2019  ·  Sarthak, Shikhar Shukla, Govind Mittal ·

Language Identification (LI) is an important first step in several speech processing systems. With a growing number of voice-based assistants, speech LI has emerged as a widely researched field. To approach the problem of identifying languages, we can either adopt an implicit approach where only the speech for a language is present or an explicit one where text is available with its corresponding transcript. This paper focuses on an implicit approach due to the absence of transcriptive data. This paper benchmarks existing models and proposes a new attention based model for language identification which uses log-Mel spectrogram images as input. We also present the effectiveness of raw waveforms as features to neural network models for LI tasks. For training and evaluation of models, we classified six languages (English, French, German, Spanish, Russian and Italian) with an accuracy of 95.4% and four languages (English, French, German, Spanish) with an accuracy of 96.3% obtained from the VoxForge dataset. This approach can further be scaled to incorporate more languages.

PDF Abstract

Datasets


Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Keyword Spotting VoxForge 2D-ConvNet Accuracy (%) 95.4 # 2
Keyword Spotting VoxForge 1D-ConvNet Accuracy (%) 93.7 # 1
Spoken language identification VoxForge Commonwealth 1D ConvNet(MixUp=NO) Accuracy (%) 93.7 # 4
Spoken language identification VoxForge Commonwealth 2D ConvNet with Attention and GRU(MixUp=YES) Accuracy (%) 95.0 # 2
Spoken language identification VoxForge Commonwealth 2D ConvNet(MixUp=YES) Accuracy (%) 95.4 # 1
Spoken language identification VoxForge Commonwealth 2D ConvNet(MixUp=NO) Accuracy (%) 94.3 # 3
Spoken language identification VoxForge European 2D ConvNet(MixUp=NO) Accuracy (%) 96.0 # 2
Spoken language identification VoxForge European 1D ConvNet(MixUp=NO) Accuracy (%) 94.4 # 4
Spoken language identification VoxForge European 2D ConvNet with Attention and GRU(MixUp=NO) Accuracy (%) 94.7 # 3
Spoken language identification VoxForge European 2D ConvNet with Attention and GRU(MixUp=YES) Accuracy (%) 93.7 # 5
Spoken language identification VoxForge European 2D ConvNet(MixUp=YES) Accuracy (%) 96.3 # 1

Methods


No methods listed for this paper. Add relevant methods here