mlsanityData to 7 Oct 2026

Learning Efficient Representations for Keyword Spotting with Triplet Loss

arXiv 2101.04792 ↗First posted 12 Jan 2021eess.AS cs.AI cs.LGcited by 24

Not on the map. Why?

Abstract (arXiv)

In the past few years, triplet loss-based metric embeddings have become a de-facto standard for several important computer vision problems, most no-tably, person reidentification. On the other hand, in the area of speech recognition the metric embeddings generated by the triplet loss are rarely used even for classification problems. We fill this gap showing that a combination of two representation learning techniques: a triplet loss-based embedding and a variant of kNN for classification instead of cross-entropy loss significantly (by 26% to 38%) improves the classification accuracy for convolutional networks on a LibriSpeech-derived LibriWords datasets. To do so, we propose a novel phonetic similarity based triplet mining approach. We also improve the current best published SOTA for Google Speech Commands dataset V1 10+2 -class classification by about 34%, achieving 98.55% accuracy, V2 10+2-class classification by about 20%, achieving 98.37% accuracy, and V2 35-class classification by over 50%, achieving 97.0% accuracy.

Key facts. Cited by 24 papers in the map (data through 7 Oct 2026). Builds on 21 of its 47 references that are in the map. First posted on arXiv 12 Jan 2021.

Builds on

21 of 47 references.

References · 21

  1. 2015FaceNet: A Unified Embedding for Face Recognition and Clusteringcited by 2,833
  2. 2017Billion-scale similarity search with GPUscited by 2,477
  3. 2017In Defense of the Triplet Loss for Person Re-Identificationcited by 784
  4. 2019Speech Model Pre-training for End-to-End Spoken Language Understandingcited by 191
  5. 2014Learning Fine-grained Image Similarity with Deep Rankingcited by 189
  6. 2017Hello Edge: Keyword Spotting on Microcontrollerscited by 108
  7. 2020Towards Learning a Universal Non-Semantic Representation of Speechcited by 84
  8. 2017Deep Residual Learning for Small-Footprint Keyword Spottingcited by 68
  9. 2017Deep Speaker: an End-to-End Neural Speaker Embedding Systemcited by 56
  10. 2020Streaming keyword spotting on mobile devicescited by 47
  11. 2018A neural attention model for speech command recognitioncited by 44
  12. 2020MatchboxNet: 1D Time-Channel Separable Convolutional Neural Network Architecture for Speech Commands Recognitioncited by 36
9 more
  1. 2017Convolutional Recurrent Neural Networks for Small-Footprint Keyword Spottingcited by 30
  2. 2017Streaming Small-Footprint Keyword Spotting using Sequence-to-Sequence Modelscited by 19
  3. 2016TristouNet: Triplet Loss for Speaker Turn Embeddingcited by 18
  4. 2017Max-Pooling Loss Training of Long Short-Term Memory Networks for Small-Footprint Keyword Spottingcited by 15
  5. 2020Metric Learning for Keyword Spottingcited by 9
  6. 2017Honk: A PyTorch Reimplementation of Convolutional Neural Networks for Keyword Spottingcited by 6
  7. 2020Howl: A Deployed, Open-Source Wake Word Detection Systemcited by 6
  8. 2018Triplet Network with Attention for Speaker Diarizationcited by 5
  9. 2020End-to-end Triplet Loss based Emotion Embedding System for Speech Emotion Recognitioncited by 4

Built on by

Cited by year

Citing papers per year: 2021: 5; 2022: 4; 2023: 7; 2024: 5; 2025: 3.

24 papers.

Sort:

By area:

  1. 2022HTS-AT: A Hierarchical Token-Semantic Audio Transformer for Sound Classification and Detectioncited by 174
  2. 2021Keyword Transformer: A Self-Attention Model for Keyword Spottingcited by 59
  3. 2024A Large-Scale Evaluation of Speech Foundation Modelscited by 38
  4. 2023SpeechPrompt v2: Prompt Tuning for Speech Classification Taskscited by 31
  5. 2023Towards Stealthy Backdoor Attacks against Speech Recognition via Elements of Soundcited by 28
  6. 2023Leveraging Large Language Models for Exploiting ASR Uncertaintycited by 15
  7. 2024LMUFormer: Low Complexity Yet Powerful Spiking Model With Legendre Memory Unitscited by 11
  8. 2023Flexible Keyword Spotting based on Homogeneous Audio-Text Embeddingcited by 10
  9. 2023Few-Shot Open-Set Learning for On-Device Customization of KeyWord Spotting Systemscited by 9
  10. 2023UniverSLU: Universal Spoken Language Understanding for Diverse Tasks with Natural Language Instructionscited by 9
  11. 2024Self-Learning for Personalized Keyword Spotting on Ultra-Low-Power Audio Sensorscited by 7
  12. 2022FedorAS: Federated Architecture Search under system heterogeneitycited by 4
12 more
  1. 2021Neural Model Reprogramming with Similarity Based Mapping for Low-Resource Spoken Command Recognitioncited by 3
  2. 2021Comparison of semi-supervised deep learning algorithms for audio classificationcited by 2
  3. 2021Attention-Free Keyword Spottingcited by 2
  4. 2022Speech Augmentation Based Unsupervised Learning for Keyword Spottingcited by 1
  5. 2023Finite Gaussian Neurons: Defending against adversarial attacks by making neural networks say "I don't know"cited by 1
  6. 2025Enhancing Few-shot Keyword Spotting Performance through Pre-Trained Self-supervised Speech Modelscited by 1
  7. 2025Language models for longitudinal analysis of abusive content in Billboard Music Chartscited by 1
  8. 2021Low-Resource Spoken Language Identification Using Self-Attentive Pooling and Deep 1D Time-Channel Separable Convolutionscited by 0
  9. 2022Learning Audio Representations with MLPscited by 0
  10. 2024Contrastive Augmentation: An Unsupervised Learning Approach for Keyword Spotting in Speech Technologycited by 0
  11. 2024TriG-NER: Triplet-Grid Framework for Discontinuous Named Entity Recognitioncited by 0
  12. 2025Edgeworth Expansion for Semi-hard Triplet Losscited by 0