High-Level CNN and Machine Learning Methods for Speaker Recognition.

Speaker Recognition (SR) is a common task in AI-based sound analysis, involving structurally different methodologies such as Deep Learning or "traditional" Machine Learning (ML). In this paper, we compared and explored the two methodologies on the DEMoS dataset consisting of 8869 audio files of 58 speakers in different emotional states. A custom CNN is compared to several pre-trained nets using image inputs of spectrograms and Cepstral-temporal (MFCC) graphs. AML approach based on acoustic feature extraction, selection and multi-class classification by means of a Naïve Bayes model is also considered. Results show how a custom, less deep CNN trained on grayscale spectrogram images obtain the most accurate results, 90.15% on grayscale spectrograms and 83.17% on colored MFCC. AlexNet provides comparable results, reaching 89.28% on spectrograms and 83.43% on MFCC.The Naïve Bayes classifier provides a 87.09% accuracy and a 0.985 average AUC while being faster to train and more interpretable. Feature selection shows how F0, MFCC and voicing-related features are the most characterizing for this SR task. The high amount of training samples and the emotional content of the DEMoS dataset better reflect a real case scenario for speaker recognition, and account for the generalization power of the models.

Artificial Intelligence-Based Voice Assessment of Patients with Parkinson's Disease Off and On Treatment: Machine vs. Deep-Learning Comparison.

Parkinson's Disease (PD) is one of the most common non-curable neurodegenerative diseases. Diagnosis is achieved clinically on the basis of different symptoms with considerable delays from the onset of neurodegenerative processes in the central nervous system. In this study, we investigated early and full-blown PD patients based on the analysis of their voice characteristics with the aid of the most commonly employed machine learning (ML) techniques. A custom dataset was made with hi-fi quality recordings of vocal tasks gathered from Italian healthy control subjects and PD patients, divided into early diagnosed, off-medication patients on the one hand, and mid-advanced patients treated with L-Dopa on the other. Following the current state-of-the-art, several ML pipelines were compared usingdifferent feature selection and classification algorithms, and deep learning was also explored with a custom CNN architecture. Results show how feature-based ML and deep learning achieve comparable results in terms of classification, with KNN, SVM and naïve Bayes classifiers performing similarly, with a slight edge for KNN. Much more evident is the predominance of CFS as the best feature selector. The selected features act as relevant vocal biomarkers capable of differentiating healthy subjects, early untreated PD patients and mid-advanced L-Dopa treated patients.

Synthetic Nuances to Maximize n-Type Organic Electrochemical Transistor and Thermoelectric Performance in Fused Lactam Polymers.

A series of fully fused n-type mixed conduction lactam polymers p(g7NCnN), systematically increasing the alkyl side chain content, are synthesized via an inexpensive, nontoxic, precious-metal-free aldol polycondensation. Employing these polymers as channel materials in organic electrochemical transistors (OECTs) affords state-of-the-art n-type performance with p(g7NC10N) recording an OECT electron mobility of 1.20 × 10-2 cm2 V-1 s-1 and a µC* figure of merit of 1.83 F cm-1 V-1 s-1. In parallel to high OECT performance, upon solution doping with (4-(1,3-dimethyl-2,3-dihydro-1H-benzoimidazol-2-yl)phenyl)dimethylamine (N-DMBI), the highest thermoelectric performance is observed for p(g7NC4N), with a maximum electrical conductivity of 7.67 S cm-1 and a power factor of 10.4 µW m-1 K-2. These results are among the highest reported for n-type polymers. Importantly, while this series of fused polylactam organic mixed ionic-electronic conductors (OMIECs) highlights that synthetic molecular design strategies to bolster OECT performance can be translated to also achieve high organic thermoelectric (OTE) performance, a nuanced synthetic approach must be used to optimize performance. Herein, we outline the performance metrics and provide new insights into the molecular design guidelines for the next generation of high-performance n-type materials for mixed conduction applications, presenting for the first time the results of a single polymer series within both OECT and OTE applications.

The Emotion Probe: On the Universality of Cross-Linguistic and Cross-Gender Speech Emotion Recognition via Machine Learning.

Machine Learning (ML) algorithms within a human-computer framework are the leading force in speech emotion recognition (SER). However, few studies explore cross-corpora aspects of SER; this work aims to explore the feasibility and characteristics of a cross-linguistic, cross-gender SER. Three ML classifiers (SVM, Naïve Bayes and MLP) are applied to acoustic features, obtained through a procedure based on Kononenko's discretization and correlation-based feature selection. The system encompasses five emotions (disgust, fear, happiness, anger and sadness), using the Emofilm database, comprised of short clips of English movies and the respective Italian and Spanish dubbed versions, for a total of 1115 annotated utterances. The results see MLP as the most effective classifier, with accuracies higher than 90% for single-language approaches, while the cross-language classifier still yields accuracies higher than 80%. The results show cross-gender tasks to be more difficult than those involving two languages, suggesting greater differences between emotions expressed by male versus female subjects than between different languages. Four feature domains, namely, RASTA, F0, MFCC and spectral energy, are algorithmically assessed as the most effective, refining existing literature and approaches based on standard sets. To our knowledge, this is one of the first studies encompassing cross-gender and cross-linguistic assessments on SER.

Deep learning and machine learning-based voice analysis for the detection of COVID-19: A proposal and comparison of architectures.

Alongside the currently used nasal swab testing, the COVID-19 pandemic situation would gain noticeable advantages from low-cost tests that are available at any-time, anywhere, at a large-scale, and with real time answers. A novel approach for COVID-19 assessment is adopted here, discriminating negative subjects versus positive or recovered subjects. The scope is to identify potential discriminating features, highlight mid and short-term effects of COVID on the voice and compare two custom algorithms. A pool of 310 subjects took part in the study; recordings were collected in a low-noise, controlled setting employing three different vocal tasks. Binary classifications followed, using two different custom algorithms. The first was based on the coupling of boosting and bagging, with an AdaBoost classifier using Random Forest learners. A feature selection process was employed for the training, identifying a subset of features acting as clinically relevant biomarkers. The other approach was centered on two custom CNN architectures applied to mel-Spectrograms, with a custom knowledge-based data augmentation. Performances, evaluated on an independent test set, were comparable: Adaboost and CNN differentiated COVID-19 positive from negative with accuracies of 100% and 95% respectively, and recovered from negative individuals with accuracies of 86.1% and 75% respectively. This study highlights the possibility to identify COVID-19 positive subjects, foreseeing a tool for on-site screening, while also considering recovered subjects and the effects of COVID-19 on the voice. The two proposed novel architectures allow for the identification of biomarkers and demonstrate the ongoing relevance of traditional ML versus deep learning in speech analysis.

Regiochemistry-Driven Organic Electrochemical Transistor Performance Enhancement in Ethylene Glycol-Functionalized Polythiophenes.

Novel p-type semiconducting polymers that can facilitate ion penetration, and operate in accumulation mode are much desired in bioelectronics. Glycol side chains have proven to be an efficient method to increase bulk electrochemical doping and optimize aqueous swelling. One early polymer which exemplifies these design approaches was p(g2T-TT), employing a bithiophene-co-thienothiophene backbone with glycol side chains in the 3,3' positions of the bithiophene repeat unit. In this paper, the analogous regioisomeric polymer, namely pgBTTT, was synthesized by relocating the glycol side chains position on the bithiophene unit of p(g2T-TT) from the 3,3' to the 4,4' positions and compared with the original p(g2T-TT). By changing the regio-positioning of the side chains, the planarizing effects of the S-O interactions were redistributed along the backbone, and the influence on the polymer's microstructure organization was investigated using grazing-incidence wide-angle X-ray scattering (GIWAXS) measurements. The newly designed pgBTTT exhibited lower backbone disorder, closer π-stacking, and higher scattering intensity in both the in-plane and out-of-plane GIWAXS measurements. The effect of the improved planarity of pgBTTT manifested as higher hole mobility (µ) of 3.44 ± 0.13 cm2 V-1 s-1. Scanning tunneling microscopy (STM) was in agreement with the GIWAXS measurements and demonstrated, for the first time, that glycol side chains can also facilitate intermolecular interdigitation analogous to that of pBTTT. Electrochemical quartz crystal microbalance with dissipation of energy (eQCM-D) measurements revealed that pgBTTT maintains a more rigid structure than p(g2T-TT) during doping, minimizing molecular packing disruption and maintaining higher hole mobility in operation mode.

Voice Analysis with Machine Learning: One Step Closer to an Objective Diagnosis of Essential Tremor.

BACKGROUND: Patients with essential tremor have upper limb postural and action tremor often associated with voice tremor. The objective of this study was to objectively examine voice tremor and its response to symptomatic pharmacological treatment in patients with essential tremor using voice analysis consisting of power spectral analysis and machine learning. METHODS: We investigated 58 patients (24 men; mean age ± SD, 71.7 ± 9.2 years; range, 38-85 years) and 74 age- and sex-matched healthy subjects (20 men; mean age ± SD, 71.0 ± 12.4 years; range, 43-95 years). We recorded voice samples during sustained vowel emission using a high-definition audio recorder. Voice samples underwent sound signal analysis, including power spectral analysis and support vector machine classification. We compared voice recordings in patients with essential tremor who did and did not manifest clinically overt voice tremor and in patients who were and were not under the symptomatic effect of the best medical treatment. RESULTS: Power spectral analysis demonstrated a prominent oscillatory activity peak at 2-6 Hz in patients who manifested a clinically overt voice tremor. Voice analysis with support vector machine classifier objectively discriminated with high accuracy between controls and patients who did and did not manifest clinically overt voice tremor and between patients who were and were not under the symptomatic effect of the best medical treatment. CONCLUSIONS: In patients with essential tremor, voice tremor is characterized by abnormal oscillatory activity at 2-6 Hz. Voice analysis, including power spectral analysis and support vector machine classification, objectively detected voice tremor and its response to symptomatic pharmacological treatment in patients with essential tremor. © 2021 International Parkinson and Movement Disorder Society.

Morphology, energetics and growth kinetics of diphenylalanine fibres.

Diphenylalanine (FF) has been shown to self-assemble from water into heterogeneous fibres that are among the stiffest biomaterials known. How and why the fibres form has, however, not been clear. In this work, the nucleation and growth of FF fibres was investigated in a combined experimental and theoretical study. Scanning electron microscopy and optical microscopy showed FF fibre morphology to be hollow tubes of varying widths with occasional endcaps. Molecular dynamics simulations of FF nanostructures based on the bulk crystalline geometry demonstrated that axial growth stablilises the fibres and that structures with different widths show similar stabilities, in accord with the wide range of fibre widths observed experimentally. Linear dichroism (LD) spectroscopy was used to determine the thermal stability of the fibres, showing that FF solutions are fully monomeric at 70 °C and that fibres begin to form at ∼40 °C upon cooling. The LD kinetic studies indicated a nucleation-driven assembly with subsequent fibre growth, but a secondary nucleation process is required to explain the data.

Machine-Learning Analysis of Voice Samples Recorded through Smartphones: The Combined Effect of Ageing and Gender.

BACKGROUND: Experimental studies using qualitative or quantitative analysis have demonstrated that the human voice progressively worsens with ageing. These studies, however, have mostly focused on specific voice features without examining their dynamic interaction. To examine the complexity of age-related changes in voice, more advanced techniques based on machine learning have been recently applied to voice recordings but only in a laboratory setting. We here recorded voice samples in a large sample of healthy subjects. To improve the ecological value of our analysis, we collected voice samples directly at home using smartphones. METHODS: 138 younger adults (65 males and 73 females, age range: 15-30) and 123 older adults (47 males and 76 females, age range: 40-85) produced a sustained emission of a vowel and a sentence. The recorded voice samples underwent a machine learning analysis through a support vector machine algorithm. RESULTS: The machine learning analysis of voice samples from both speech tasks discriminated between younger and older adults, and between males and females, with high statistical accuracy. CONCLUSIONS: By recording voice samples through smartphones in an ecological setting, we demonstrated the combined effect of age and gender on voice. Our machine learning analysis demonstrates the effect of ageing on voice.

The Effect of Ring Expansion in Thienobenzo[b]indacenodithiophene Polymers for Organic Field-Effect Transistors.

A fused donor, thienobenzo[b]indacenodithiophene (TBIDT), was designed and synthesized using a novel acid-promoted cascade ring closure strategy, and then copolymerized with a benzothiadiazole (BT) monomer. The backbone of TBIDT is an expansion of the well-known indacenodithiophene (IDT) unit and was expected to enhance the charge carrier mobility by improving backbone planarity and facilitating short contacts between polymer chains. However, the optimized field-effect transistors demonstrated an average saturation hole mobility of 0.9 cm2 V-1 s-1, lower than the performance of IDT-BT (∼1.5 cm2 V-1 s-1). Mobilities extracted from time-resolved microwave conductivity measurements were consistent with the trend in hole mobilities in organic field-effect transistor devices. Scanning tunneling microscopy measurements and computational modeling illustrated that TBIDT-BT exhibits a less ordered microstructure in comparison to IDT-BT. This reveals that a regular side-chain packing density, independent of conformational isomers, is critical to avoid local free volume due to irregular packing, which can host trapping impurities. DFT calculations indicated that TBIDT-BT, despite containing a larger, planar unit, showed less stabilization of planar backbone geometries in comparison to IDT-BT. This is due to the reduced electrostatic stabilizing interactions between the peripheral thiophene of the fused core and the BT unit, resulting in a reduction of the barrier to rotation around the single bond. These insights provide a greater understanding of the general structure-property relationships required for semiconducting polymer repeat units to ensure optimal backbone planarization, as illustrated with IDT-type units, guiding the design of novel semiconducting polymers with extended fused backbones for high-performance field-effect transistors.

TCNQ Physisorption on the Topological Insulator Bi2 Se3.

Topological insulators are promising candidates for spintronic applications due to their topologically protected, spin-momentum locked and gapless surface states. The breaking of the time-reversal symmetry after the introduction of magnetic impurities, such as 3d transition metal atoms embedded in two-dimensional molecular networks, could lead to several phenomena interesting for device fabrication. The first step towards the fabrication of metal-organic coordination networks on the surface of a topological insulator is to investigate the adsorption of the pure molecular layer, which is the aim of this study. Here, the effect of the deposition of the electron acceptor 7,7,8,8-tetracyanoquinodimethane (TCNQ) molecules on the surface of a prototypical topological insulator, bismuth selenide (Bi2 Se3 ), is investigated. Scanning tunneling microscope images at low-temperature reveal the formation of a highly ordered two-dimensional molecular network. The essentially unperturbed electronic structure of the topological insulator observed by photoemission spectroscopy measurements demonstrates a negligible charge transfer between the molecular layer and the substrate. Density functional theory calculations confirm the picture of a weakly interacting adsorbed molecular layer. These results reveal significant potential of TCNQ for the realization of metal-organic coordination networks on the topological insulator surface.

Two-Dimensional Ketone-Driven Metal-Organic Coordination on Cu(111).

Two-dimensional metal-organic nanostructures based on the binding of ketone groups and metal atoms were fabricated by depositing pyrene-4,5,9,10-tetraone (PTO) molecules on a Cu(111) surface. The strongly electronegative ketone moieties bind to either copper adatoms from the substrate or codeposited iron atoms. In the former case, scanning tunnelling microscopy images reveal the development of an extended metal-organic supramolecular structure. Each copper adatom coordinates to two ketone ligands of two neighbouring PTO molecules, forming chains that are linked together into large islands through secondary van der Waals interactions. Deposition of iron atoms leads to a transformation of this assembly resulting from the substitution of the metal centres. Density functional theory calculations reveal that the driving force for the metal substitution is primarily determined by the strength of the ketone-metal bond, which is higher for Fe than for Cu. This second class of nanostructures displays a structural dependence on the rate of iron deposition.

Os2 -Os4 Switch Controls DNA Knotting and Anticancer Activity.

Dinuclear trihydroxido-bridged osmium-arene complexes are inert and biologically inactive, but we show here that linking dihydroxido-bridged Os(II) -arene fragments by a bridging di-imine to form a metallacycle framework results in strong antiproliferative activity towards cancer cells and distinctive knotting of DNA. The shortened spacer length reduces biological activity and stability in solution towards decomposition to biologically inactive dimers. Significant differences in behavior toward plasmid DNA condensation are correlated with biological activity.

The synthesis and STM/AFM imaging of 'olympicene' benzo[cd]pyrenes.

H-Benzo[cd]pyrene ('Olympicene') is a polyaromatic hydrocarbon and non-Kekulé fragment of graphene. A new synthetic method has been developed for the formation of 6H-benzo[cd]pyrene and related ketones including the first time isolation of the unstable alcohol 6H-benzo[cd]pyren-6-ol. Molecular imaging of the reaction products with scanning tunnelling microscopy (STM) and non-contact atomic force microscopy (NC-AFM) characterised the 6H-benzo[cd]pyrene as well as the previously intangible and significantly less stable 5H-benzo[cd]pyrene, the fully conjugated benzo[cd]pyrenyl radical and the ketones as oxidation products.

Self-assembly of decoupled borazines on metal surfaces: the role of the peripheral groups.

Two borazine derivatives have been synthesised to investigate their self-assembly behaviour on Au(111) and Cu(111) surfaces by scanning tunnelling microscopy (STM) and theoretical simulations. Both borazines form extended 2D networks upon adsorption on both substrates at room temperature. Whereas the more compact triphenyl borazine 1 arranges into close-packed ordered molecular islands with an extremely low density of defects on both substrates, the tris(phenyl-4-phenylethynyl) derivative 2 assembles into porous molecular networks due to its longer lateral substituents. For both species, the steric hindrance between the phenyl and mesityl substituents results in an effective decoupling of the central borazine core from the surface. For borazine 1, this is enough to weaken the molecule-substrate interaction, so that the assemblies are only driven by attractive van der Waals intermolecular forces. For the longer and more flexible borazine 2, a stronger molecule-substrate interaction becomes possible through its peripheral substituents on the more reactive copper surface.

Molecular-Scale Imaging Enables Direct Visualization of Molecular Defects and Chain Structure of Conjugated Polymers.

Conjugated polymers have become materials of choice for applications ranging from flexible optoelectronics to neuromorphic computing, but their polydispersity and tendency to aggregate pose severe challenges to their precise characterization. Here, the combination of vacuum electrospray deposition (ESD) with scanning tunneling microscopy (STM) is used to acquire, within the same experiment, assembly patterns, full mass distributions, exact sequencing, and quantification of polymerization defects. In a first step, the ESD-STM results are successfully benchmarked against NMR for low molecular mass polymers, where this technique is still applicable. Then, it is shown that ESD-STM is capable of reaching beyond its limits by characterizing, with the same accuracy, samples that are inaccessible to NMR. Finally, a recalibration procedure is proposed for size exclusion chromatography (SEC) mass distributions, using ESD-STM results as a reference. The distinctiveness of the molecular-scale information obtained by ESD-STM highlights its role as a crucial technique for the characterization of conjugated polymers.

The Role of Side Chains and Hydration on Mixed Charge Transport in n-Type Polymer Films.

Introducing ethylene glycol (EG) side chains to a conjugated polymer backbone is a well-established synthetic strategy for designing organic mixed ion-electron conductors (OMIECs). However, the impact that film swelling has on mixed conduction properties has yet to be scoped, particularly for electron-transporting (n-type) OMIECs. Here, the authors investigate the effect of the length of branched EG chains on mixed charge transport of n-type OMIECs based on a naphthalene-1,4,5,8-tetracarboxylic-diimide-bithiophene backbone. Atomic force microscopy (AFM), grazing-incidence wide-angle X-ray scattering (GIWAXS), and scanning tunneling microscopy (STM) are used to establish the similarities between the common-backbone films in dry conditions. Electrochemical quartz crystal microbalance with dissipation monitoring (EQCM-D) and in situ GIWAXS measurements reveal stark changes in film swelling properties and microstructure during electrochemical doping, depending on the side chain length. It is found that even in the loss of the crystallite content upon contact with the aqueous electrolyte, the films can effectively transport charges and that it is rather the high water content that harms the electronic interconnectivity within the OMIEC films. These results highlight the importance of controlling water uptake in the films to impede charge transport in n-type electrochemical devices.

Fibrillation of transferrin.

BACKGROUND: The nature of fibrillar deposits from aqueous solutions of human serum and recombinant human transferrin on mica and carbon-coated formvar surfaces has been investigated. METHODS AND RESULTS: Atomic force microscopy showed that the deposition of recombinant transferrin onto the hydrophilic surface of mica resulted in the formation of a monolayer-thick film composed of conformationally-strained flattened protein molecules. Elongated fibres developed on top of this layer and appeared to be composed of single proteins or small clusters thereof. Monomeric and dimeric transferrins were separated by gel permeation chromatography and their states of aggregation confirmed by mass spectrometry and dynamic light scattering. Transmission electron-microscopy showed that dimeric transferrin, but not monomeric transferrin, deposited on carbon-coated formvar grids forms rounded (circular) structures ca. 250nm in diameter. Small transferrin fibrils ca. 250nm long appeared to be composed of smaller rounded sub-units. Synchrotron radiation-circular dichroism and, Congo red and thioflavin-T dye-binding experiments suggested that transferrin aggregation in solution does not involve major structural changes to the protein or formation of classical ß-sheet amyloid structures. Collisional cross sections determined via ion mobility-mass spectrometry showed little difference between the overall protein shapes of apo- and holo-transferrin in the gas phase. GENERAL SIGNIFICANCE: The possibility that transferrin deformation and aggregation are involved in neurological disorders such as Parkinson's and Alzheimer's disease is discussed. This article is part of a Special Issue entitled Transferrins: Molecular mechanisms of iron transport and disorders.