PCA and deep learning based myoelectric grasping control of a prosthetic hand

Background: For the functional control of prosthetic hand, it is insufficient to obtain only the motion pattern information. As far as practicality is concerned, the control of the prosthetic hand force is indispensable. The application value of prosthetic hand will be greatly improved if the stable grip of prosthetic hand can be achieved. To address this problem, in this study, a bio-signal control method for grasping control of a prosthetic hand is proposed to improve patient's sense of using prosthetic hand and the thus improving the quality of life. Methods: A MYO gesture control armband is used to collect the surface electromyographic (sEMG) signals from the upper limb. The overlapping sliding window scheme are applied for data segmentation and the correlated features are extracted from each segmented data. Principal component analysis (PCA) methods are then deployed for dimension reduction. Deep neural network is used to generate sEMG-force regression model for force prediction at different levels. The predicted force values are input to a fuzzy controller for the grasping control of a prosthetic hand. A vibration feedback device is used to feed grasping force value back to patient's arm to improve patient's sense of using prosthetic hand and realize accurate grasping. To test the effectiveness of the scheme, 15 able-bodied subjects participated in the experiments. Results: The classification results indicated that 8-channel sEMG applying all four time-domain features, with PCA reduction from 32 to 8 dimensions results in the highest classification accuracy. Based on the experimental results from 15 participants, the average recognition rate is over 95%. On the other hand, from the statistical results of standard deviation, the between-subject variations ranges from 3.58 to 1.25%, proving that the robustness and stability of the proposed approach. Conclusions: The method proposed hereto control grasping power through the patient's own sEMG signal, which achieves a high recognition rate to improve the success rate of grip and increases the sense of operation and also brings the gospel for upper extremity amputation patients.

Sparse and Low-rank Modeling for Automatic Speech Recognition

Pranay Dighe

This thesis deals with exploiting the low-dimensional multi-subspace structure of speech towards the goal of improving acoustic modeling for automatic speech recognition (ASR). Leveraging the parsimonious hierarchical nature of speech, we hypothesize that whenever a speech signal is measured in a high-dimensional feature space, the true class information is embedded in low-dimensional subspaces whereas noise is scattered as random high-dimensional erroneous estimations in the features. In this context, the contribution of this thesis is twofold: (i) identify sparse and low-rank modeling approaches as excellent tools for extracting the class-specific low-dimensional subspaces in speech features, and (ii) employ these tools under novel ASR frameworks to enrich the acoustic information present in the speech features towards the goal of improving ASR. Techniques developed in this thesis focus on deep neural network (DNN) based posterior features which, under the sparse and low-rank modeling approaches, unveil the underlying class-specific low-dimensional subspaces very elegantly. In this thesis, we tackle ASR tasks of varying difficulty, ranging from isolated word recognition (IWR) and connected digit recognition (CDR) to large-vocabulary continuous speech recognition (LVCSR). For IWR and CDR, we propose a novel \textit{Compressive Sensing} (CS) perspective towards ASR. Here exemplar-based speech recognition is posed as a problem of recovering sparse high-dimensional word representations from compressed low-dimensional phonetic representations. In the context of LVCSR, this thesis argues that albeit their power in representation learning, DNN based acoustic models still have room for improvement in exploiting the \textit{union of low-dimensional subspaces} structure of speech data. Therefore, this thesis proposes to enhance DNN posteriors by projecting them onto the manifolds of the underlying classes using principal component analysis (PCA) or compressive sensing based dictionaries. Projected posteriors are shown to be more accurate training targets for learning better acoustic models, resulting in improved ASR performance. The proposed approach is evaluated on both close-talk and far-field conditions, confirming the importance of sparse and low-rank modeling of speech in building a robust ASR framework. Finally, the conclusions of this thesis are further consolidated by an information theoretic analysis approach which explicitly quantifies the contribution of proposed techniques in improving ASR.

EPFL2019

Sparse and Low-rank Modeling for Automatic Speech Recognition

Pranay Dighe

EPFL2019