Speech enhancement in flat monitor TVs

Speech enhancement in flat monitor TVs The further development of technology is leading to wider and thinner flat TV sets. This trend is beneficial in terms of ease of installation and area occupation but very poor for audio quality. Thinner TVs impose loudspeakers to be sized extremely small causing a worsen in audio performance. Speech intelligibility is not immune to this problem and final users often complain about it. The phenomenon can arise in different situations: for example, when the broadcasting company do not optimize the audio for the TV, or when recordings have poor quality. Several solutions have been proposed in recent years, each preferring the development of additional hardware (like a soundbar to be placed under the TV). However, most of the final users just rely on the built-in loudspeakers of the appliance and don't have any audio dedicated equipment. The best approach to solve the intelligibility problem would be developing a custom equalization strategy for each TV in every possible configuration, but it is practically unfeasible. Therefore a new method has been conceived with the aim of being universal, able to work with different models of TV sets in typical indoor environment. The method consists in a dynamic equalization which enhances a precise frequency range, without heavily altering the original signal. The algorithm was tested using several broadcast materials and two different methods: subjective and objective evaluations. This project was financed by RAI, the major broadcast company of Italy, which promoted a research activity named \textit{"Fornitura di un sistema per l’analisi e il miglioramento della qualità audio percepita da TV a schermo piatto"}, contract number 1183102135/00, date 12/11/2018. The project aims at developing a trasfer function (f.d.t.), to be implemented on a Digital Audio Optimizer, able to improve the listening experience of the customer regarding the quality of the perceived audio on the majority of available TV sets (dimensions between 37'' and 55''). The algorithm must include an objective evaluation method, implemented on a workstation, to evaluate the influence of the algorithm and consequently compute the expected subjective grade. In other words, the audio must be evaluated with and without the algorithm through a dedicated software. The goal is an improvement between the audio quality of the signal processed and the signal not processed by at least 20%. The subjective test involved more than 70 participants and obtained an average 25% improvement, thus meeting the goal. The objective test was organized according to a combination of acoustical and psycho-acoustical parameters and the correlation between the subjective test was computed.