Sparse Autoencoder-Based Feature Transfer Learning for Speech Emotion Recognition

Abstract

Abstract-In speech emotion recognition, training and test data used for system development usually tend to fit each other perfectly, but further 'similar' data may be available. Transfer learning helps to exploit such similar data for training despite the inherent dissimilarities in order to boost a recogniser's performance. In this context, this paper presents a sparse autoencoder method for feature transfer learning for speech emotion recognition. In our proposed method, a common emotion-specific mapping rule is learnt from a small set of labelled data in a target domain. Then, newly reconstructed data are obtained by applying this rule on the emotion-specific data in a different domain. The experimental results evaluated on six standard databases show that our approach significantly improves the performance relative to learning each source domain independently.

A survey of transfer learning

“…In the paper by Romera-Paredes [97], a multi-task transfer learning approach is used to predict pain levels from an individual's facial expression by using labeled source facial images from other individuals. The paper by Deng [25] applies transfer learning to the application of speech emotion recognition where information is transferred from multiple labeled speech sources. The application of wine quality classification is implemented in Zhang [141] using a multi-task transfer learning approach.…”

Perspectives and Future Outlook of Deep Learning AI

“…This has a connection to itself at the next time step that has a weight of one, so it copies its own real-valued state and accumulates the external signal, but this self-connection is multiplicatively gated by another unit that learns to decide when to clear the content of the memory. RNNs are very powerful dynamic systems, but training them has proved to be problematic because the back propagated gradients either grow or shrink at each time step leading to vanish or exploding gradient problems [20]. RNNs can be viewed as very deep feed forward networks in which all the layers share the same weights.…”

“…Hochreiter et al [20] in their work on LSTMs describe the error back flow problems solved by the use of the LSTM. This paper has demonstrated that character-level speech transcription can be performed by a recurrent neural network with minimal preprocessing and no explicit phonetic representation.…”

WUB-IP: A High-Precision UWB Positioning Scheme for Indoor Multiuser Applications

“…Transfer learning was proposed to reuse the knowledge learned previously from other data [20], [21]. The idea is to utilize commonalities among different learning tasks to share statistical strength, and transfer the common knowledge across tasks [21]- [23]. In this research, the sharedhidden-layer autoencoder (SHLA) is utilized to extract the common knowledge among different conditions, and realize transfer learning.…”

“…In order to improve the ranging accuracy under NLOS conditions, a transfer learning-based approach is adopted to mitigate the ranging error. Transfer learning was proposed to reuse the knowledge learned previously from other data [20], [21]. The idea is to utilize commonalities among different learning tasks to share statistical strength, and transfer the common knowledge across tasks [21]- [23].…”