Refine
Year of publication
Document Type
- Conference Proceeding (14)
- Article (6)
- Part of a Book (2)
- Book (1)
- Doctoral Thesis (1)
- Patent (1)
Keywords
- Düngemittel (1)
- Landwirtschaft (1)
- Maschinelles Lernen (1)
- Maschinelles Sehen (1)
- Spracherkennung (1)
Department/institution
As the recognition of emotion from speech has matured to a degree where it becomes applicable in real-life settings, it is time for a realistic view on obtainable performances. Most studies tend to overestimation in this respect: acted data is often used rather than spontaneous data, results are reported on pre-selected prototypical data, and true speaker disjunctive partitioning is still less common than simple cross-validation. A considerably more realistic impression can be gathered by inter-set evaluation: we therefore show results employing six standard databases in a cross-corpora evaluation experiment. To better cope with the observed high variances, different types of normalization are investigated. 1.8k individual evaluations in total indicate the crucial performance inferiority of inter- to intra-corpus testing.
Sensing and acquiring reliable physical values are the fundamentals, not only for a predictive maintenance or quality assessment, but especially for big data analysis and sophisticated Industry 4.0 applications. In steel plants, physical values are distributed over the complete process chain of steel making while the environmental conditions are harsh with respect to high temperature, aggressive fluids, water, shock and dust. These conditions render the use of electronic devices focusing a consumer market impossible.The developed self-contained smart measuring system presented in this paper survives in harsh environments and is composed of small-sized modules providing miscellaneous functionalities. The high degree of modularity in hard-and software facilitates a cost-effective adaptation to many applications, like vibration monitoring, temperature logging or torque measurement. Different onboard measurement components are available yet, namely high sensitive bridge amplifiers for strain measurements, accelerometers, gyroscopes, orientation-, temperature- and humidity sensors as well as an unique system identification. A communication module enables wireless transmission of the acquired data via Bluetooth or NFC online. Moreover, different power supply features are supported by the power module: Ultra-low power modes for long-life battery use, recharging of lithium cells, and an inductive power supply for wireless power transfer for applications with moving or rotating components. The heart of the system is a powerful ARM based microcontroller which enables an intelligent analysis of the data in situ. This is especially important where data size and complexity is the relevant factor for example in area-wide sensor networks.
Deep Neural Networks (DNNs) denote multilayer artificial neural networks with more than one hidden layer and millions of free parameters. We propose a Generalized Discriminant Analysis (GerDA) based on DNNs to learn discriminative features of low dimension optimized with respect to a fast classification from a large set of acoustic features for emotion recognition. On nine frequently used emotional speech corpora, we compare the performance of GerDA features and their subsequent linear classification with previously reported benchmarks obtained using the same set of acoustic features classified by Support Vector Machines (SVMs). Our results impressively show that low-dimensional GerDA features capture hidden information from the acoustic features leading to a significantly raised unweighted average recall and considerably raised weighted average recall.