Te Meng Ting’s research while affiliated with University of Science Malaysia and other places

This paper shows how two microphones in an endfire array configuration was used to perform beamforming. The setup uses two condenser microphones and a sound card to allow multiple sources to be input to the computer at the same time. A cross-correlation calculation was used to determine the time shift between the two mics. Using the Delay and Sum algorithm, the time shift can be corrected, and the mic signals can be added to a superposition.

Material classification is pivotal across materials science, engineering, and various industrial sectors. Despite the high accuracy of traditional material classification methods, they often entail large, intricate, and costly setups that demand skilled operators. In this study, we introduce the MCT-array, a newly developed compact RF antenna array system measuring 100×100×2mm, which functions as a transceiver. This device, equipped with 32 receiving antennas and 2 transmitters, leverages dynamic power adjustments to refine material detection accuracy. The study evaluates three machine learning classifiers, namely Support Vector Machine (SVM), Multi-Layer Perceptron (MLP), and Random Forest (RandF) on twelve different materials. MATLAB simulations are initially conducted to identify optimal transceiver configurations. Following the identification of optimal parameters from these simulations, real-world experiments are conducted with the materials positioned 30 cm away from the antenna. Results demonstrate that RandF achieves a material classification accuracy of 94.84%, followed by SVM at 94.5%, and MLP at 94.1%. Detailed analysis further reveals that RandF is the preferred option for tasks demanding the highest levels of accuracy, SVM strikes an optimal balance between processing speed and accuracy, while MLP stands out for its rapid prediction times, making it especially suitable for real-time applications. Integrating an innovative portable RF transceiver with these machine learning models, achieving an impressive average accuracy of over 94%, represents a scalable and effective solution. This innovation holds significant promise for sectors engaged in material classification, particularly in the realms of robotics and automation.

In this work, a binaural model resembling the human auditory system was built using a pair of three-dimensional (3D)-printed ears to localize a sound source in both vertical and horizontal directions. An analysis on the proposed model was firstly conducted to study the correlations between the spatial auditory cues and the 3D polar coordinate of the source. Apart from the estimation techniques via interaural and spectral cues, the property from the combined direct and reverberant energy decay curve is also introduced as part of the localization strategy. The preliminary analysis reveals that the latter provides a much more accurate distance estimation when compared to approximations via sound pressure level approach, but is alone not sufficient to disambiguate the front-rear confusions. For vertical localization, it is also shown that the elevation angle can be robustly encoded through the spectral notches. By analysing the strengths and shortcomings of each estimation method, a new algorithm is formulated to localize the sound source which is also further improved by cross-correlating the interaural and spectral cues. The proposed technique has been validated via a series of experiments where the sound source was randomly placed at 30 different locations in an outdoor environment up to a distance of 19 m. Based on the experimental and numerical evaluations, the localization performance has been significantly improved with an average error of 0.5 m from the distance estimation and a considerable reduction of total ambiguous points to 3.3%.