Classification of Phonetic Syllables Using Stacked Autoencoder and Characterization via Centroid
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
This work presents a new method for layer insertion in stacked autoencoder neural networks. In this approach, a branch of layers is inserted on the side of the last hidden layer, and the output layer after the training of an existing layer has stabilized. Later, the new branch is merged with the previous layers. This insertion type is called collaborative, as it introduces new knowledge to the network without reducing the knowledge already acquired by the previous layers. This approach enables a neural network to learn with reduced design time. It overcomes the typical problem of defining the number of layers and the number of neurons in each layer. This technique was applied in phonetic syllable classification, where the Fast Fourier Transform obtains the audio data. These audio data were processed using a vertical bar plot to compress the audio data by using centroids. This procedure provided data compression without losing characteristics. Thus, collaborative insertions were evaluated in terms of the degree of growth for a multi-class classification problem.
Fields of Study & Subjects
Neural Networks and Applications
Speech Recognition and Synthesis
Machine Learning and ELM
Publication details
- Journal / source
- International Conference on Artificial Intelligence in Information and Communication (ICAIIC)
- Pages
- 744–749
- Publication date
- 24 February 2026
- Language
- EN