Fig 2 - uploaded by Huy Phan
Content may be subject to copyright.
Download
View publication
Illustration of the processing pipeline to train the CNN for deep feature embedding learning.

Illustration of the processing pipeline to train the CNN for deep feature embedding learning.

Source publication
Fig. 1. The two-level hierarchy of scene categories constructed based...
Fig. 2. Illustration of the processing pipeline to train the CNN for...
Fig. 3. Illustration of extracting high-level features from the learned...
Fig. 4. Category-wise performance comparison between the proposed...
Deep Feature Embedding and Hierarchical Classification for Audio Scene Classification
Preprint
Full-text available
  • Feb 2020
In this work, we propose an approach that features deep feature embedding learning and hierarchical classification with triplet loss function for Acoustic Scene Classification (ASC). In the one hand, a deep convolutional neural network is firstly trained to learn a feature embedding from scene audio signals. Via the trained convolutional neural net...
Cite
Download full-text

Contexts in source publication

Context 1
... processing pipeline for deep feature embedding learning using a deep CNN is illustrated in Fig. 2. Each acoustic scene signal is firstly transformed into time-frequency image, such as Gammatone spectrogram with 128 Gammatone filters [11]. The time-frequency image is then decomposed into nonoverlapping image patches of size 128 × 128. Let X and y denote an image patch and its one-hot encoding label, respectively. Mixup data ...
Context 2
... clarity, in Fig. 2 and Table I, we intentionally separate the deep CNN into two parts: the CNN part for feature learning ...
Context 3
... processing pipeline for deep feature embedding learning using a deep CNN is illustrated in Fig. 2. Each acoustic scene signal is firstly transformed into time-frequency image, such as Gammatone spectrogram with 128 Gammatone filters [11]. The time-frequency image is then decomposed into nonoverlapping image patches of size 128 × 128. Let X and y denote an image patch and its one-hot encoding label, respectively. Mixup data ...
Context 4
... clarity, in Fig. 2 and Table I, we intentionally separate the deep CNN into two parts: the CNN part for feature learning ...

Similar publications

Figure 1: Our Proposed Solution
Figure 3: Architecture of the Hierarchical Attention Network
Comparison of results
F1-scores for each class
A Novel Multimodal Music Genre Classifier using Hierarchical Attention and Convolutional Neural Network
Preprint
Full-text available
  • Nov 2020
Music genre classification is one of the trending topics in regards to the current Music Information Retrieval (MIR) Research. Since, the dependency of genre is not only limited to the audio profile, we also make use of textual content provided as lyrics of the corresponding song. We implemented a CNN based feature extractor for spectrograms in ord...
View
Figure 2: Self-similarity matrix computed from the SAE embeddings. The...
Self-Expressing Autoencoders for Unsupervised Spoken Term Discovery
Preprint
Full-text available
  • Jul 2020
Unsupervised spoken term discovery consists of two tasks: finding the acoustic segment boundaries and labeling acoustically similar segments with the same labels. We perform segmentation based on the assumption that the frame feature vectors are more similar within a segment than across the segments. Therefore, for strong segmentation performance,...
View
FIGURE 1. The framework of our proposed method.
FIGURE 2. Data augmentation.
FIGURE 6. T-SNE visualization of emotion embedding on IEMOCAP (10-fold).
Autoencoder With Emotion Embedding for Speech Emotion Recognition
Article
Full-text available
  • Mar 2021
An important part of the human-computer interaction process is speech emotion recognition (SER), which has been receiving more attention in recent years. However, although a wide diversity of methods has been proposed in SER, these approaches still cannot improve the performance. A key issue in the low performance of the SER system is how to effect...
View
Recognition rate of five models on the data set.
Research on Traffic Acoustic Event Detection Algorithm Based on Sparse Autoencoder
Article
Full-text available
  • Jan 2020
Road traffic monitoring is very important for intelligent transportation. The detection of traffic state based on acoustic information is a new research direction. A vehicles acoustic event classification algorithm based on sparse autoencoder is proposed to analysis the traffic state. Firstly, the multidimensional Mel-cepstrum features and energy f...
View