Machine learning for decoding listeners' attention from electroencephalography evoked by continuous speech.

Previous research has shown that it is possible to predict which speaker is attended in a multispeaker scene by analyzing a listener's electroencephalography (EEG) activity. In this study, existing linear models that learn the mapping from neural activity to an attended speech envelope are replaced by a non-linear neural network (NN). The proposed architecture takes into account the temporal context of the estimated envelope and is evaluated using EEG data obtained from 20 normal-hearing listeners who focused on one speaker in a two-speaker setting. The network is optimized with respect to the frequency range and the temporal segmentation of the EEG input, as well as the cost function used to estimate the model parameters. To identify the salient cues involved in auditory attention, a relevance algorithm is applied that highlights the electrode signals most important for attention decoding. In contrast to linear approaches, the NN profits from a wider EEG frequency range (1-32 Hz) and achieves a performance seven times higher than the linear baseline. Relevant EEG activations following the speech stimulus after 170 ms at physiologically plausible locations were found. This was not observed when the model was trained on the unattended speaker. Our findings therefore indicate that non-linear NNs can provide insight into physiological processes by analyzing EEG activity.

Single-ended prediction of listening effort using deep neural networks.

The effort required to listen to and understand noisy speech is an important factor in the evaluation of noise reduction schemes. This paper introduces a model for Listening Effort prediction from Acoustic Parameters (LEAP). The model is based on methods from automatic speech recognition, specifically on performance measures that quantify the degradation of phoneme posteriorgrams produced by a deep neural net: Noise or artifacts introduced by speech enhancement often result in a temporal smearing of phoneme representations, which is measured by comparison of phoneme vectors. This procedure does not require a priori knowledge about the processed speech, and is therefore single-ended. The proposed model was evaluated using three datasets of noisy speech signals with listening effort ratings obtained from normal hearing and hearing impaired subjects. The prediction quality was compared to several baseline models such as the ITU-T standard P.563 for single-ended speech quality assessment, the American National Standard ANIQUE+ for single-ended speech quality assessment, and a single-ended SNR estimator. In all three datasets, the proposed new model achieved clearly better prediction accuracies than the baseline models; correlations with subjective ratings were above 0.9. So far, the model is trained on the specific noise types used in the evaluation. Future work will be concerned with overcoming this limitation by training the model on a variety of different noise types in a multi-condition way in order to make it generalize to unknown noise types.

Porphyrin Metalation at MgO Surfaces: A Spectroscopic and Quantum Mechanical Study on Complementary Model Systems.

We show that both single-crystalline and nanostructured MgO surfaces convert free-base tetraphenyl porphyrin (2HTPP) into magnesium tetraphenyl porphyrin (MgTPP) at room temperature. The reaction can be viewed as an ion exchange between the two aminic protons of the 2HTPP molecule with a Mg(2+) ion from the surface. The driving force for the reaction is the strong stability of the formed hydroxyl groups along the steps and at defects on the MgO surface. We have used an integrated characterization approach that includes UV/Vis diffuse reflectance measurements on nanostructured powders, X-ray photoelectron spectroscopic investigation of atomically clean MgO(100) single-crystalline thin films, and density functional theory (DFT) calculations on model systems. The DFT calculations demonstrate that MgTPP formation is strongly exothermic at the corners, edges and steps, but slightly endothermic on terrace sites. This agrees well with the UV/Vis diffuse reflectance, which upon adsorption of 2HTPP shows a decrease in the absorption band associated with corner and edge sites on MgO nanocube powders.

NMR for direct determination of K(m) and V(max) of enzyme reactions based on the Lambert W function-analysis of progress curves.

(1)H NMR spectroscopy was used to follow the cleavage of sucrose by invertase. The parameters of the enzyme's kinetics, K(m) and V(max), were directly determined from progress curves at only one concentration of the substrate. For comparison with the classical Michaelis-Menten analysis, the reaction progress was also monitored at various initial concentrations of 3.5 to 41.8mM. Using the Lambert W function the parameters K(m) and V(max) were fitted to obtain the experimental progress curve and resulted in K(m)=28mM and V(max)=13µM/s. The result is almost identical to an initial rate analysis that, however, costs much more time and experimental effort. The effect of product inhibition was also investigated. Furthermore, we analyzed a much more complex reaction, the conversion of farnesyl diphosphate into (+)-germacrene D by the enzyme germacrene D synthase, yielding K(m)=379µM and k(cat)=0.04s(-1). The reaction involves an amphiphilic substrate forming micelles and a water insoluble product; using proper controls, the conversion can well be analyzed by the progress curve approach using the Lambert W function.

Effect of speech-intrinsic variations on human and automatic recognition of spoken phonemes.

The aim of this study is to quantify the gap between the recognition performance of human listeners and an automatic speech recognition (ASR) system with special focus on intrinsic variations of speech, such as speaking rate and effort, altered pitch, and the presence of dialect and accent. Second, it is investigated if the most common ASR features contain all information required to recognize speech in noisy environments by using resynthesized ASR features in listening experiments. For the phoneme recognition task, the ASR system achieved the human performance level only when the signal-to-noise ratio (SNR) was increased by 15 dB, which is an estimate for the human-machine gap in terms of the SNR. The major part of this gap is attributed to the feature extraction stage, since human listeners achieve comparable recognition scores when the SNR difference between unaltered and resynthesized utterances is 10 dB. Intrinsic variabilities result in strong increases of error rates, both in human speech recognition (HSR) and ASR (with a relative increase of up to 120%). An analysis of phoneme duration and recognition rates indicates that human listeners are better able to identify temporal cues than the machine at low SNRs, which suggests incorporating information about the temporal dynamics of speech into ASR systems.

Peptide NMHRYPNQ of the cellular prion protein (PrP(C)) inhibits aggregation and is a potential key for understanding prion-prion interactions.

Pathogenesis of transmissible spongiform encephalopathies is correlated with a conversion of the normal cellular form of the prion protein (PrP(C)) into the abnormal isoform (scrapie form of PrP). Contact of the normal PrP with its abnormal isoform, the scrapie form of PrP, induces the transformation. Knowledge of molecules that inhibit such contacts leads to an understanding of the mechanism of the aggregation, and these molecules may serve as leads for drugs against transmissible spongiform encephalopathies. Therefore, we screened a synthetic octapeptide library of the globular domain of the human PrP(C) for binding affinity to PrP(C). Two fragments with binding affinity, 149YYRENMHR156 and 153NMHRYPNQ160, were identified with K(d) values of 21 and 25 microM, respectively. A 10-fold excess of peptide 153NMHRYPNQ160 inhibits aggregation of the PrP by 99%. NMR and mass spectrometry showed that the binding region of the peptide 153NMHRYPNQ160 is located at helix 3 of the PrP.

Perspectives on NMR in drug discovery: a technique comes of age.

In the past decade, the potential of harnessing the ability of nuclear magnetic resonance (NMR) spectroscopy to monitor intermolecular interactions as a tool for drug discovery has been increasingly appreciated in academia and industry. In this Perspective, we highlight some of the major applications of NMR in drug discovery, focusing on hit and lead generation, and provide a critical analysis of its current and potential utility.

Phenylthiazolyl-hydrazide and its derivatives are potent inhibitors of tau aggregation and toxicity in vitro and in cells.

One of the key pathological features of Alzheimer's disease is the aggregation of tau protein. We are therefore searching for compounds capable of inhibiting this reaction. On the basis of an initial screen of 200000 compounds [Pickhardt, M., Gazova, Z., von Bergen, M., Khlistunova, I., Wang, Y., Hascher, A., Mandelkow, E. M., Biernat, J., and Mandelkow, E. (2005) Anthraquinones inhibit tau aggregation and dissolve Alzheimer's paired helical filaments in vitro and in cells, J. Biol. Chem. 280, 3628-3635], we performed an in silico screen and predicted a new phenylthiazolyl-hydrazide (PTH) compound as a possible hit [Larbig, G., Pickhardt, M., Lloyd, D. G., Schmidt, B., and Mandelkow, E. (2007) Screening for inhibitors of tau protein aggregation into Alzheimer paired helical filaments: A ligand based approach results in successful scaffold hopping. Curr. Alzheimer Res. 4 (3), 315-323.]. Synthesis of this compound showed that it was indeed active in terms of inhibiting de novo tau aggregation and disassembling preformed aggregates (IC50 = 7.7 microM and DC50 = 10.8 microM). We have now synthesized 49 similar structures and identified the core of the PTHs to be crucial for activity, thus representing a lead structure. Analysis of the binding epitope by saturation transfer difference NMR shows strong interactions between the tau protein and the ligand in the aromatic regions of the inhibitor. By chemical variation of the core, we improved the inhibitory potency five-fold. The compounds showed a low toxicity as judged by an N2A cell model of tau aggregation and lend themselves for further development.