RECORD
FrAUG: A Frame Rate Based Data Augmentation Method for Depression Detection from Speech Signals
Audio and Speech Processing · Submitted 11 Feb 2022
abstract compiled · 2026-09-23 · no language-model review
The authors' main claim Further, the proposed method outperformed commonly used data-augmentation methods such as noise augmentation, VTLP, Speed, and Pitch Perturbation.
The compiled record
The paper's own sentences, arranged. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. Numbers in the tables are the authors', checked by code. No sentence in this record was written by a model.
Open the compiled record
The abstract, sentence by sentence
In this paper, a data augmentation method is proposed for depression detection from speech signals. Samples for data augmentation were created by changing the frame-width and the frame-shift parameters during the feature extraction process. Unlike other data augmentation methods (such as VTLP, pitch perturbation, or speed perturbation), the proposed method does not explicitly change acoustic parameters but rather the time-frequency resolution of frame-level features. The proposed method was evaluated using two different datasets, models, and input acoustic features. For the DAIC-WOZ (English) dataset when using the DepAudioNet model and mel-Spectrograms as input, the proposed method resulted in an improvement of 5.97% (validation) and 25.13% (test) when compared to the baseline. The improvements for the CONVERGE (Mandarin) dataset when using the x-vector embeddings with CNN as the backend and MFCCs as input features were 9.32% (validation) and 12.99% (test). Baseline systems do not incorporate any data augmentation. Further, the proposed method outperformed commonly used data-augmentation methods such as noise augmentation, VTLP, Speed, and Pitch Perturbation. All improvements were statistically significant.
background contribution method result interpretation main claim
| # | sentence | role | the authors' own work | states the main conclusion | strength |
|---|---|---|---|---|---|
| s0 | In this paper, a data augmentation method is proposed for depression detection from speech signals. | contribution 1.00 | 0.83 | 0.06 | asserted 0.99 |
| s1 | Samples for data augmentation were created by changing the frame-width and the frame-shift parameters during the feature extraction process. | method 1.00 | 0.94 | 0.03 | asserted 1.00 |
| s2 | Unlike other data augmentation methods (such as VTLP, pitch perturbation, or speed perturbation), the proposed method does not explicitly change acoustic parameters but rather the time-frequency resolution of frame-level features. | contribution 0.55 | 0.42 | 0.06 | asserted 1.00 |
| s3 | The proposed method was evaluated using two different datasets, models, and input acoustic features. | method 0.91 | 0.93 | 0.03 | asserted 1.00 |
| s4 | For the DAIC-WOZ (English) dataset when using the DepAudioNet model and mel-Spectrograms as input, the proposed method resulted in an improvement of 5.97% (validation) and 25.13% (test) when compared to the baseline. | result 1.00 | 0.97 | 0.19 | asserted 1.00 |
| s5 | The improvements for the CONVERGE (Mandarin) dataset when using the x-vector embeddings with CNN as the backend and MFCCs as input features were 9.32% (validation) and 12.99% (test). | result 1.00 | 0.95 | 0.10 | asserted 1.00 |
| s6 | Baseline systems do not incorporate any data augmentation. | background 0.45 | 0.34 | 0.03 | asserted 1.00 |
| s7 | Further, the proposed method outperformed commonly used data-augmentation methods such as noise augmentation, VTLP, Speed, and Pitch Perturbation. | result 1.00 | 0.88 | 0.31 | asserted 1.00 |
| s8 | All improvements were statistically significant. | result 0.85 | 0.91 | 0.15 | asserted 1.00 |
What was examined, and what was not
- arXiv listing abstract v1
- full text
- figures and tables
- references and citing sentences
- relations to other papers
- P abstract
- S sections
- T body sentences
- C citations
- R cross-paper relations
The paper
Original abstract
In this paper, a data augmentation method is proposed for depression detection from speech signals. Samples for data augmentation were created by changing the frame-width and the frame-shift parameters during the feature extraction process. Unlike other data augmentation methods (such as VTLP, pitch perturbation, or speed perturbation), the proposed method does not explicitly change acoustic parameters but rather the time-frequency resolution of frame-level features. The proposed method was evaluated using two different datasets, models, and input acoustic features. For the DAIC-WOZ (English) dataset when using the DepAudioNet model and mel-Spectrograms as input, the proposed method resulted in an improvement of 5.97% (validation) and 25.13% (test) when compared to the baseline. The improvements for the CONVERGE (Mandarin) dataset when using the x-vector embeddings with CNN as the backend and MFCCs as input features were 9.32% (validation) and 12.99% (test). Baseline systems do not incorporate any data augmentation. Further, the proposed method outperformed commonly used data-augmentation methods such as noise augmentation, VTLP, Speed, and Pitch Perturbation. All improvements were statistically significant.
Human review
How the literature uses it
2 arXiv papers cite this in 2 sentences
Papers citing it · 1 sentence without a confident label
Cites it as background 1 paper
-
Aside from these handcrafted features, some approaches use raw audio signals [ref] in an end-to-end manner or high-level features extracted from pre-trained models, like x-vector [ref] , i-vector [ref] , NN2Vec [ref] , and Depression audio embedding (DEPA) [ref] .
Unsupervised Instance Discriminative Learning for Depression Detection from Speech Signals · Introduction · 0.88
What it cites · 4 sentences without a confident label
Uses the method or tool of 1 paper
-
The MUSAN library [ref] was used to augment every utterance with randomly chosen foreground noise samples at a randomly chosen SNR of 0,5,10, or 15 dB [ref] .
MUSAN: A Music, Speech, and Noise Corpus · FrAUG versus Conventional Data Augmentation Methods · 0.93
Compares with 2 papers
-
This is comparable to the reported F1-scores of 0.610 in prior works [ref] .
Gender Bias in Depression Detection Using Audio Features · Multi Frame Rate Training · 0.99
-
This best performing configuration is better than any of the single frame rate performances, including when only the frame-widths are manipulated as in [ref] .
Multi-Window Data Augmentation Approach for Speech Emotion Recognition · Multi Frame Rate Training · 0.99
Cites as background 1 paper
-
In [ref] , a multi-window data augmentation was proposed for emotion recognition which used multiple frame-widths.
Multi-Window Data Augmentation Approach for Speech Emotion Recognition · Introduction · 0.80
Discussion
Know something about this paper? Leave a comment, or highlight any passage above to comment on it.