Advanced algorithms for audio quality improvement in musical keyboards instruments

The focus of this thesis is on the signal processing techniques used to increase the audio quality of the most common digital audio effects employed in electronic musical instruments also taking into account the feasibility of the proposed algorithms’ implementation in accordance with the design constraints and the available computational limits. More in detail four different issues have been analyzed throughout this dissertation: artificial reverberation, analysis and emulation of nonlinear devices, audio morphing, and room response equalization. Among the audio effects, one of the most used is definitely artificial reverberation. A great deal of research has been devoted in the last decades to improve the performance of digital artificial reverberators. Thanks to the progress of technology the traditional techniques based on recursive structures (i.e., IIR filters) are accompanied by new approaches based on fast convolution techniques and hybrid reverberator structures. On this basis, an efficient real-time implementation of a fast convolution algorithm has been proposed taking into account an embedded system. Moreover, a technique for reducing the computational load required by this operation using psychoacoustic expedients has been presented considering a joint assessment of the energy decay relief and the absolute threshold of hearing. Finally, some techniques for the approximation of the convolution operation with recursive structures at low computational cost have been suggested. Doctoral School on Engineering Sciences Universita Politecnica delle Marche Although the convolution operation allows the exact reproduction of a linear system, it is important to consider that most of the audio effects are nonlinear systems (i.e., compressors, distortion, amplifiers). For this reason, the most commonly used techniques for the emulation of nonlinear systems based on a black box approach have been studied and analyzed. In particular, a technique for the approximation of the dynamic convolution operation by exploiting the principal component analysis has been proposed. Using this procedure it is possible to reduce the cost of dynamic convolution without lowering the perceived audio quality. An adaptive algorithm for the identification of nonlinear systems using orthogonal functions has also been presented. In order to provide greater flexibility and major artistic expression to musicians, several audio morphing techniques have been analyzed. In particular, this procedure makes possible to combine two or more audio signals in order to create new sounds that are acoustically interesting. This study has led to the development of an audio morphing algorithm for percussive hybrid sound generation. The main features of the presented approach are preprocessing of the audio references performed in the frequency domain and time domain linear interpolation to execute the morphing. Finally, equalization techniques for improving the quality of sound reproduction systems by compensating the room transfer function have been taken into account. In particular, two algorithms for adaptive minimum-phase equalization and a mixed-phase equalization technique have been proposed. In order to verify the suitability of the proposed systems, experiments on a realistic scenario have been carried out.