Automatic selection of spoken language biomarkers for dementia detection.

This paper analyzes diverse features extracted from spoken language to select the most discriminative ones for dementia detection. We present a two-step feature selection (FS) approach: Step 1 utilizes filter methods to pre-screen features, and Step 2 uses a novel feature ranking (FR) method, referred to as dual dropout ranking (DDR), to rank the screened features and select spoken language biomarkers. The proposed DDR is based on a dual-net architecture that separates FS and dementia detection into two neural networks (namely, the operator and selector). The operator is trained on features obtained from the selector to reduce classification or regression loss. The selector is optimized to predict the operator's performance based on automatic regularization. Results show that the approach significantly reduces feature dimensionality while identifying small feature subsets that achieve comparable or superior performance compared with the full, default feature set. The Python codes are available at https://github.com/kexquan/dual-dropout-ranking.

Speaker-turn aware diarization for speech-based cognitive assessments.

Introduction: Speaker diarization is an essential preprocessing step for diagnosing cognitive impairments from speech-based Montreal cognitive assessments (MoCA). Methods: This paper proposes three enhancements to the conventional speaker diarization methods for such assessments. The enhancements tackle the challenges of diarizing MoCA recordings on two fronts. First, multi-scale channel interdependence speaker embedding is used as the front-end speaker representation for overcoming the acoustic mismatch caused by far-field microphones. Specifically, a squeeze-and-excitation (SE) unit and channel-dependent attention are added to Res2Net blocks for multi-scale feature aggregation. Second, a sequence comparison approach with a holistic view of the whole conversation is applied to measure the similarity of short speech segments in the conversation, which results in a speaker-turn aware scoring matrix for the subsequent clustering step. Third, to further enhance the diarization performance, we propose incorporating a pairwise similarity measure so that the speaker-turn aware scoring matrix contains both local and global information across the segments. Results: Evaluations on an interactive MoCA dataset show that the proposed enhancements lead to a diarization system that outperforms the conventional x-vector/PLDA systems under language-, age-, and microphone-mismatch scenarios. Discussion: The results also show that the proposed enhancements can help hypothesize the speaker-turn timestamps, making the diarization method amendable to datasets without timestamp information.

Underwater Acoustic Target Recognition Based on Supervised Feature-Separation Algorithm.

For the purpose of improving the accuracy of underwater acoustic target recognition with only a small number of labeled data, we proposed a novel recognition method, including 4 steps: pre-processing, pre-training, fine-tuning and recognition. The 4 steps can be explained as follows: (1) Pre-processing with Resonance-based Sparsity Signal Decomposition (RSSD): RSSD was firstly utilized to extract high-resonance components from ship-radiated noise. The high-resonance components contain the major information for target recognition. (2) Pre-training with unsupervised feature-extraction: we proposed a one-dimensional convolution autoencoder-decoder model and then we pre-trained the model to extract features from the high-resonance components. (3) Fine-tuning with supervised feature-separation: a supervised feature-separation algorithm was proposed to fine-tune the model and separate the extracted features. (4) Recognition: classifiers were trained to recognize the separated features and complete the recognition mission. The unsupervised pre-training autoencoder-decoder can make good use of a large number of unlabeled data, so that only a small number of labeled data are required in the following supervised fine-tuning and recognition, which is quite effective when it is difficult to collect enough labeled data. The recognition experiments were all conducted on ship-radiated noise data recorded using a sensory hydrophone. By combining the 4 steps above, the proposed recognition method can achieve recognition accuracy of 93.28%, which sufficiently surpasses other traditional state-of-art feature-extraction methods.