Distributed speech recognition (DSR) splits the processing of data between a mobile device and a network server. In the front-end, features are extracted and compressed to transmit over a wireless channel to a back-end server, where the incoming stream is received and reconstructed for recognition tasks. In this paper, we propose a feature compression algorithm termed suppression by selecting wavelets (SSW) for DSR: minimizing memory and device requirements while also maintaining or even improving the recognition performance. The SSW approach first applies the discrete wavelet transform (DWT) to filter the incoming speech feature sequence into two temporal sub-sequences at the client terminal. Feature compression is achieved by keeping the low (modulation) frequency sub-sequence while discarding the high frequency counterpart. The low-frequency sub-sequence is then transmitted across the remote network for specific feature statistics normalization. Wavelets are favorable for resolving the temporal properties of the feature sequence, and the down-sampling process in DWT reduces the amount of data at the terminal prior to transmission across the network, which can be interpreted as data compression. Once the compressed features have arrived at the server, the feature sequence can be enhanced by statistics normalization, reconstructed with inverse DWT, and compensated with a simple post filter to alleviate any over-smoothing effects from the compression stage. Results on a standard robustness task (Aurora-4) and on a Mandarin Chinese news corpus (MATBN) showed SSW outperforms conventional noise-robustness techniques while also providing nearly a 50% compression rate during the transmission stage of DSR systems.

Conventionally, the maximum likelihood (ML) criterion is applied to train a deep belief network (DBN). We present a maximum entropy (ME) learning algorithm for DBNs, designed specifically to handle limited training data. Maximizing only the entropy of parameters in the DBN allows more effective generalization capability, less bias towards data distributions, and robustness to over-fitting compared to ML learning. Results of text classification and object recognition tasks demonstrate ME-trained DBN outperforms ML-trained DBN when training data is limited.

Implantable microelectrodes are useful for monitoring neural response patterns in the auditory cortex, however chronic neural recordings can often deteriorate with time (e.g. impedance measures across electrode arrays generally increase monotonically over the first 7 days post-implant). This problem is caused by the increasing spatial distribution of reactive tissue responses (corresponding to changes in impedance spectra along the electrode-tissue-interface). Therefore, the design of microelectrode probes must ensure that the neuronal ensembles lie within a cylindrical radius of the recording electrodes. In this chapter, chronic neural recording failure is examined via cortical spike patterns, histological analyses, indentation experiments, and finite element models. Next, the microfabrication of the “Utah” electrode array and the “Michigan” probe is compared to determine how their size, shape, and geometry address: (1) the spatial distribution of neurons (as related to recording quality); (2) the initial penetrating profile (as related to insertion killzones); (3) the reactive cell responses (as related to glial encapsulation); (4) the anchoring of the probe’s position in the tissue (as related to micromotions) and (5) the embedding of various bioactive reagents (ex: growth factors, anti-inflammatory drugs, etc.). Finally, a novel hydrogel “Dropping Method” is proposed for controlling the biocompatibility and mechanical properties at the electrode-tissue-interface.

Temporal Filterbanks in Cochlear Implant Hearing and Deep Learning Simulations | InTechOpen, Published on: 2017-03-29. Authors: Payton Lin