Neural Network Architecture That Combines Temporal and Summative Features for Infant Cry Classification in the Interspeech 2018 Computational Paralinguistics Challenge

Abstract

This paper describes the application of a novel deep neural network architecture to the classification of infant vocalisations as part of the Interspeech 2018 Computational Paralinguistics Challenge. Previous approaches to infant cry classification have either applied a statistical classifier to summative features of the whole cry, or applied a syntactic pattern recognition technique to a temporal sequence of features. In this work we explore a deep neural network architecture that exploits both temporal and summative features to make a joint classification. The temporal input comprises centi-second frames of low-level signal features which are input to LSTM nodes, while the summative vector comprises a large set of statistical functionals of the same frames that are input to MLP nodes. The combined network is jointly optimized and evaluated using leave-one-speaker-out cross-validation on the challenge training set. Results are compared to independently-trained temporal and summative networks and to a baseline SVM classifier. The combined model outperforms the other models and the challenge baseline on the training set. While problems remain in finding the best configuration and training protocol for such networks, the approach seems promising for future signal classification tasks.

Comparison of Artificial Neural Network Types for Infant Vocalization Classification

“…This result reflects recent developments in the DCASE community, in which virtually all competition systems contain convolutional stages [21], [22], [25], [26], [28], [43]- [46]. However the paralinguistics community still relies primarily on pure recurrent networks [8], [9], [18], [19].…”

“…Recurrent Stage: Gated recurrent units (GRU) and long short-term memory units (LSTM) [54] are the two most common recurrent types in paralinguistics [9], [18], [19], [55], [56]. Unidirectional [9], [18] as well as bidirectional [19], [56] networks are popular. We used the CuDNN implementations 3 to reduce the training time of the recurrent units.…”

“…There are also Intermediate approaches between these two types, which typically calculate hand-crafted feature sets on audio frames (so called "low-level descriptors"(LLDs)) and input the resulting feature maps to RNNs [9]. Zhang et al [18] and Huckvale et al [19] reached 70.1 % UAR and 68.28 % UAR respectively through this approach. Wagner et al [9] directly dedicated their submission to the debate of end-toend vs conventional systems by comparing the performance of feature maps with varying complexity for RNNs.…”

A review of infant cry analysis and classification

“…Our previous work [15] has shown that combining weighted prosodic features with MFCC features help improve the classification accuracy in a deep learning model. Other researchers have also found that F0 is critical in identifying infant cry signals [40]. Chittora and Patil used F0 to calculate unvoiced segments ratio and found out unvoiced percentage in a cry is an important parameter for analysis of infant cry [19].…”

“…The most popular probabilistic classifier used in infant cry classification is Support Vector Machine (SVM) [26,40,43]. Many machine learning methods have been experimented in infant research.…”