The creation of Personal Sound Zones (PSZ) is a recent application of digital signal processing that allows differentiating the sound intensity in neighboring regions of space (e.g. a “bright” and a “dark” zone). Given the impulse responses of the environment, digital filters can be designed in order to obtain an attenuation of the signal in the dark zone as a result of the superposition of the filtered IR coming from each loudspeaker. A neural optimization approach was recently shown to enable PSZ by designing digital FIR filters. In this work we propose an improvement of that neural optimization approach using a simpler neural network architecture. Furthermore we extend the method to the design of IIR filters, which is computationally more effective for a real-time implementation. The neural technique is compared with two state-of-the-art methods, analyzing the performance in terms of Acoustic Contrast. Experiments have been performed using a vehicle composed of standard loudspeakers and two speaker arrays, and show that the proposed approach achieves remarkable Acoustic Contrast without sacrificing audio quality.

In this paper, a performance analysis on the estimation of the so-called observation filter for the Virtual Microphone Technique (VMT) in a realistic automotive environment is presented. A performance comparison between adaptive and fixed observation filter estimation methods, namely Least Mean Square (LMS) and Minimum Mean Square Error (MMSE), respectively, was carried on. Two different experimental setups were implemented on a popular B-segment car. Eight microphones were placed at the monitoring and virtual positions in order to sense environmental acoustic noise propagating within the cabin of the car running at variable speed on a smooth asphalt. Our experimental results show that a large spectral coherence between monitoring and virtual microphone signals indicates a potentially effective and relatively wide-band virtual microphone signal reconstruction. The fixed observation filter estimation method achieves better performance than the adaptive one, guaranteeing remarkable broadband estimation accuracy. Moreover, for each considered setup, design guidelines are proposed to obtain a good trade-off between estimation accuracy and material costs.

This paper describes a novel Deep Learning method for the design of IIR parametric filters for automatic multipoint audio equalization, that is the task of improving the sound quality of a listening environment at multiple listening points employing multiple loudspeakers. The filters are designed to approximate the inverse of the RIR and achieve almost flat magnitude response. A simple and effective neural architecture, named BiasNet, is proposed to determine the IIR equalizer parameters. This novel architecture is conceived for optimization and, as such, is able to produce optimal IIR equalizer parameters at its output, after training, with no input required. In absence of input, the presence of learnable non-zero bias terms ensures that the network works properly. An output scaling method is used to obtain accurate tuning of the IIR filters center frequency, quality factor and gain. All layers involved in the proposed method are shown to be differentiable, allowing backpropagation to optimize the network weights and achieve, after a number of training iterations, the optimal output according to a given RIR. The parameters are optimized with respect to a loss function based on a spectral distance between the measured and desired magnitude response, and a regularization term is used to keep the same microphone-loudspeaker energy balance after equalization. Two experimental scenarios are employed, a room and a car cabin, with several loudspeakers. The performance of the proposed method improves over the baseline techniques and achieves an almost flat band at a lower computational cost.

In this paper, we present an innovative inertial navigation system based on the data collected through the Inertial Measurement Unit (IMU) embedded in a commercial smart-phone. We propose an innovative step detection algorithm which is independent of the holding mode, the only assumption being that the device is hand-held (i.e., the user is texting/navigating or phoning) and its movement is related to the upper body displacement during walking. We also present a new approach able to automatically calibrate the step length estimation formula according to the smartphone positioning. The developed algorithms have been validated through a test campaign in which we have evaluated the system performance considering three different smartphone models and different path lengths. The obtained results show that the maximum step detection error is always below 4% (average: 2.08%; standard deviation: 1.82%) whereas the maximum path length estimation error is below 8.1% (average: 3.6%; standard deviation: 1.81%) in all the considered cases.

In this paper, we compare two novel algorithms for pedestrian navigation based on signals collected by a single wearable Magnetic, Angular Rate, and Gravity (MARG) sensor. The two navigation algorithms, denoted as Enhanced Pedestrian Dead Reckoning (EPDR) and De-Drifted Propagation (DDP), require the placement of the MARG sensor on the foot or on the chest of the test subject, respectively. Different methods for gait characterization are compared, evaluating navigation dynamics by using data collected through an extensive experimental campaign. The main goal of this research is to investigate the peculiarities of different inertial navigation algorithms, in order to highlight the impact of the sensor's placement, together with inertial sensor issues. Considering a closed path (i.e., ending at the starting point), the relative distance error between the starting point and the final estimated position is about 2% of the total travelled distance for both DDP and EPDR navigation algorithms. On the other hand, the error between the initial heading angle and the final estimated one is approximately 10° for EPDR and 7° for DDP, respectively.

This paper describes the development of a multifloor hybrid inertial/barometric navigation system. The prototype integrates several Magnetic Angular Rate and Gyroscope (MARG) sensors and a barometer. The inertial (MARG) sub-system, by properly processing the signals collected by the MARG sensors placed on the test subject's feet, reconstructs the two-dimensional navigation pattern by applying a Zero velocity UPdaTe (ZUPT) technique. Three different sensors' configurations are investigated in order to find the best performing set-up. A simpler configuration with a single MARG sensor is also considered to derive a reference performance benchmark without multiple MARG sensor fusion. The barometer, connected via usb to a Freakduino board, is used to detect the floor change. The fusion of inertial and barometric signals allows to fully reconstruct the movement of a person in both indoor and outdoor environments. The main goal of the proposed system is to allow accurate personal navigation without any external reference (i.e., radio signals, satellite signals, etc.). Considering a closed path, the relative distance error between the starting point and the final estimated position is below 2.5% of the total traveled distance.