# Improvements to deep convolutional neural networks for LVCSR

arXiv: https://arxiv.org/abs/1309.1501 · HTML: https://mlsanity.com/paper/1309.1501/

## Key facts

- Cited by 11 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 5 Sep 2013.

## Built on by (most cited first, 11 of 11)

- [Adan: Adaptive Nesterov Momentum Algorithm for Faster Optimizing Deep Models](https://mlsanity.com/paper/2208.06677/) (2022)
- [Branchformer: Parallel MLP-Attention Architectures to Capture Local and Global Context for Speech Recognition and Understanding](https://mlsanity.com/paper/2207.02971/) (2022)
- [Watch or Listen: Robust Audio-Visual Speech Recognition with Visual Corruption Modeling and Reliability Scoring](https://mlsanity.com/paper/2303.08536/) (2023)
- [A Comparative Study on E-Branchformer vs Conformer in Speech Recognition, Translation, and Understanding Tasks](https://mlsanity.com/paper/2305.11073/) (2023)
- [Revisiting and Exploring Efficient Fast Adversarial Training via LAW: Lipschitz Regularization and Auto Weight Averaging](https://mlsanity.com/paper/2308.11443/) (2023)
- [Expressivity and Approximation Properties of Deep Neural Networks with ReLUk Activation](https://mlsanity.com/paper/2312.16483/) (2023)
- [Improving Medical Speech-to-Text Accuracy with Vision-Language Pre-training Model](https://mlsanity.com/paper/2303.00091/) (2023)
- [MixRep: Hidden Representation Mixup for Low-Resource Speech Recognition](https://mlsanity.com/paper/2310.18450/) (2023)
- [Tab2vox: CNN-Based Multivariate Multilevel Demand Forecasting Framework by Tabular-To-Voxel Image Conversion](https://mlsanity.com/paper/2209.10516/) (2022)
- [Statistics and Deep Learning-based Hybrid Model for Interpretable Anomaly Detection](https://mlsanity.com/paper/2202.12720/) (2022)
- [Hybrid CNN-Transformer Architecture for Arabic Speech Emotion Recognition](https://mlsanity.com/paper/2604.07357/) (2026)

## Abstract (arXiv)

Deep Convolutional Neural Networks (CNNs) are more powerful than Deep Neural Networks (DNN), as they are able to better reduce spectral variation in the input signal. This has also been confirmed experimentally, with CNNs showing improvements in word error rate (WER) between 4-12% relative compared to DNNs across a variety of LVCSR tasks. In this paper, we describe different methods to further improve CNN performance. First, we conduct a deep analysis comparing limited weight sharing and full weight sharing with state-of-the-art features. Second, we apply various pooling strategies that have shown improvements in computer vision to an LVCSR speech task. Third, we introduce a method to effectively incorporate speaker adaptation, namely fMLLR, into log-mel features. Fourth, we introduce an effective strategy to use dropout during Hessian-free sequence training. We find that with these improvements, particularly with fMLLR and dropout, we are able to achieve an additional 2-3% relative improvement in WER on a 50-hour Broadcast News task over our previous best CNN baseline. On a larger 400-hour BN task, we find an additional 4-5% relative improvement over our previous best CNN baseline.