· 2015
Audio augmentation for speech recognition
- 1,155citations
- 2015year
Short summary
Changing audio speed by factors of 0.9, 1.0, and 1.1 improved large-vocabulary speech recognition (LVCSR) performance by an average of 4.3% across four tasks with 100-1000 hours of training data.
AI-generated from the title and abstract; the full text is not read.
Key points
- Audio speed perturbation (factors 0.9, 1.0, 1.1) is proposed as an audio-level data augmentation method.
- The technique has a low implementation cost and is easy to adopt.
- An average relative improvement of 4.3% was observed across four LVCSR tasks.
- Effectiveness was examined across training data sizes ranging from 100 to 1000 hours.
AI-generated from the title and abstract; the full text is not read.
Abstract
Data augmentation is a common strategy adopted to increase the quantity of training data, avoid overfitting and improve robustness of the models. In this paper, we investigate audio-level speech augmentation methods which directly process the raw signal. The method we particularly recommend is to change the speed of the audio signal, producing 3 versions of the original signal with speed factors of 0.9, 1.0 and 1.1. The proposed technique has a low implementation cost, making it easy to adopt. We present results on 4 different LVCSR tasks with training data ranging from 100 hours to 1000 hours, to examine the effectiveness of audio augmentation in a variety of data scenarios. An average relative improvement of 4.3% was observed across the 4 tasks.
The authors' abstract, as published at the source., 2015 · DOI ↗
Continue with a free account
Ask the paper: 3 free questions a day about this paper; save it, get its citation, new summaries every day for your field. Takeaways are Premium.
Continue free on the webSign in with Google or Apple; no card needed. You come back to this paper.
On your phone:
Field: Signal Processing
Signal ProcessingComputer Science