Search

Conventional direction-of-arrival (DOA) estimation algorithms for shallow water environments usually contain high amounts of error due to the presence of many acoustic reflective surfaces and scattering fields. Utilizing data from a single acoustic vector sensor, the magnitude and DOA of an acoustic signature can be estimated; as such, DOA algorithms are used to reduce the error in these estimations. Three experiments were conducted using a moving boat as an acoustic target in a waterway in Houghton, Michigan. The shallow and narrow waterway is a complex and non-linear environment for DOA estimation. This paper compares minimizing DOA errors using conventional and machine learning algorithms. The conventional algorithm uses frequency-masking averaging, and the machine learning algorithms incorporate two recurrent neural network architectures, one shallow and one deep network. Results show that the deep neural network models the shallow water environment better than the shallow neural network, and both networks are superior in performance to the frequency-masking average method. [ABSTRACT FROM AUTHOR]

Vehicle interior noise is a quality criterion of passenger cars. A considerable amount of resources is used to evaluate and design the acoustic environment with respect to given requirements. The customer's perception in the end-of-line vehicle is the main criterion. Therefore, full vehicle testing is a large part of today's sound comfort development. To increase efficiency, it is desirable to limit the hardware testing to a specific component. A later reassembly of the full vehicle is done virtually using transfer functions. These transfer functions of the substructures can be derived numerically or through measurements. However, full vehicle simulations are still challenging. Hence, transfer functions are typically measured but come with the burden of complex procedures. In this work, the authors propose a machine learning algorithm to reduce the effort for finding suitable transfer models in the automotive context. Artificial neural networks with rectified linear unit and swish activation functions are trained on full vehicle measurements. Multiple operation conditions are used for training. The networks compute spectral system responses and relative sensitivities for the input features. The performance is discussed with respect to the full vehicle validation data. The results indicate an effective procedure to reduce the costs of full-size vehicle measurements. [ABSTRACT FROM AUTHOR]