Christian H. Kasess

Effect of train type on annoyance and acoustic features of the rolling noise

This study investigated the annoyance associated with the rolling noise of different railway stock. Passbys of nine train types (passenger and freight trains) equipped with different braking systems were recorded. Acoustic features showed a clear distinction of the braking system with the A-weighted energy equivalent sound level (LAeq) showing a difference in the range of 10 dB between cast-iron braked trains and trains with disk or K-block brakes. Further, annoyance was evaluated in a psychoacoustic experiment where listeners rated the relative annoyance of the rolling noise for the different train types. Stimuli with and without the original LAeq differences were tested. For the original LAeq differences, the braking system significantly affected the annoyance with cast-iron brakes being most annoying, most likely as a consequence of the increased wheel roughness causing an increased LAeq. Contribution of the acoustic features to the annoyance was investigated revealing that the LAeq explained up to 94% of the variance. For the stimuli without differences in the LAeq, cast-iron braked train types were significantly less annoying and the spectral features explained up to 60% of the variance in the annoyance. The effect of these spectral features on the annoyance of the rolling noise is discussed.

Time-frequency multitaper methods applied to rumble strip noise

Spectral estimation techniques are important tools in different fields of acoustics such as speech processing or traffic noise. In traffic noise typically conventional techniques are used such as third octave band filterbanks and long term averages. Certain situations where e.g. the acoustical events are short and have strong tonal components may, however, require more sophisticated methods of analysis. An example of such a situation are rumble strips which are a safety measure to keep cars from running off the road. Rumble strips typically consist of grooves milled at regular intervals next to the outermost traffic lanes, i.e. on the hard shoulder. Thus, when a vehicle passes over the grooves a signal with strong tonal components is produced which should warn the driver from the imminent danger of running of the road. Here, the potential of different multitaper methods is investigated using a recording of a car passing over such a rumble strip. The main interest is whether the tonal components can be reliably detected. In multitaper approaches the time-frequency analysis is performed using different orthogonal window functions. The time-frequency representations for each window are then combined to form a single estimate of the spectrum. To detect the tonal parts, different approaches using either the conventional multitaper method as well as combining multitaper with reassignment or synchrosqueezing are investigated. The comparison is done using either simple combinations of the modified spectrograms or applying random projections to the multiple short-term Fourier transforms before the reassignment/synchrosqueezing as recently suggested. The results illustrate that the different methods applied in this work are well suited for the analysis of the rumble strip data.

The relation between psychoacoustical factors and annoyance under different noise reduction conditions for railway noise

The A-weighted sound pressure level (SPL) is commonly used to assess the effect of noise reduction measures on noise-induced annoyance. While for road traffic noise loudness seems to be a better descriptor of annoyance, for railway noise a systematic investigation seems to be lacking. Thus, in this study, the relation between annoyance and perceptually motivated descriptors was investigated for various conditions of binaural recordings of pass-bys of cargo and passenger trains. The conditions included free field and spectral mitigations caused by a 4 m high noise barrier, a 1 m high noise barrier close to the track, and rail dampers. Forty listeners performed a free magnitude estimation of annoyance for different presentation levels and the ratings were fit to various models. Further, level changes required to evoke a noticeable change in annoyance (annoyance thresholds) were acquired. The models based on the A-weighted SPL explained the ratings and thresholds better when the reduction measure was explicitly provided as a parameter. However, the optimal models were loudness-level-based models, which were able to better describe the annoyance, even independently of the reduction measure. Both experiments underline the effectiveness of loudness when describing the annoyance in the area of railway noise reduction.

Alveolar and Velarized Laterals in Albanian and in the Viennese Dialect

A comparison of alveolar and velarized lateral realizations in two language varieties, Albanian and the Viennese dialect, has been performed. Albanian distinguishes the two laterals phonemically, whereas in the Viennese dialect, the velarized lateral was introduced by language contact with Czech immigrants. A categorical distinction between the two lateral phonemes is fully maintained in Albanian. Results are not as straightforward in the Viennese dialect. Most prominently, female speakers, if at all, realize the velarized lateral in word-final position, thus indicating the application of a phonetically motivated process. The realization of the velarized lateral by male speakers, on the other hand, indicates that the velarized lateral replaced the former alveolar lateral phoneme. Alveolar laterals are either realized in perceptually salient positions, thus governed by an input-switch rule, or in front vowel contexts, thus subject to coarticulatory influences. Our results illustrate the subtle interplay of phonology, phonetics and sociolinguistics.

Bayesian vocal tract model estimates of nasal stops for speaker verification

In this paper we report on speaker verification experiments using branched vocal tract model estimates of alveolar nasal (/n/) stops. While the discriminatory potential of nasal acoustics has long been established, their acoustic properties have so far mostly been characterized using spectral features. Here, we used a Bayesian estimation technique to obtain reflection coefficients of a branched-tube model of the combined nasal and oral tract. Parameters were then modeled using probabilistic linear discriminant analysis to calculate likelihood ratios for speaker verification trials. Performance was assessed on normal and high vocal effort speech using high-quality and mobile-telephone-transmitted recordings taken from the German-language Pool2010 corpus. Results are compared with those of systems based on mel-frequency cepstral coefficients (MFCC). Vocal tract parameter based systems outperform MFCC based systems in matched conditions, but lack robustness under mismatch, while being readily interpretable with respect to a physical speech production model.

Experiments on using vocal tract estimates of nasal stops for speaker verification

Nasal stops have been recognized as an important source of speaker-discriminating features. The nasal cavity is, with the exception of the velar junction, independent of articulatory movements. As the complex nasal structure varies from person to person, features dependent upon nasal acoustics may have low within-speaker and high between-speaker variability. In this study we use a Bayesian estimation technique to obtain reflection coefficients of a branched-tube model of the combined nasal and oral tract. These are then used as parameters in speaker verification experiments. The performance is evaluated on the basis of speakers from the TIMIT corpus as well as the Kiel corpus and is compared with that of a system based on Mel frequency cepstral coefficient (MFCC) features. Fusion of both systems indicates that the two approaches offer complementary information.