Gender Identification in a Two-Level Hierarchical Speech Emotion Recognition System for an Italian Social Robot

Abstract

The real challenge in Human-Robot Interaction (HRI) is to build machines capable of perceiving human emotions so that robots can interact with humans in a proper manner. Emotion varies accordingly to many factors, and gender represents one of the most influential ones: an appropriate gender-dependent emotion recognition system is recommended indeed. In this article, we propose a Gender Recognition (GR) module for the gender identification of the speaker, as a preliminary step for the final development of a Speech Emotion Recognition (SER) system. The system was designed to be installed on social robots for hospitalized and living at home patients monitoring. Hence, the importance of reducing the software computational effort of the architecture also minimizing the hardware bulkiness, in order for the system to be suitable for social robots. The algorithm was executed on the Raspberry Pi hardware. For the training, the Italian emotional database EMOVO was used. Results show a GR accuracy value of 97.8%, comparable with the ones found in the literature.

Acoustic Gender and Age Classification as an Aid to Human–Computer Interaction in a Smart Home Environment

“…The best system achieved a 2.0% SAD error rate at the frame level, which is already usable in real-life scenarios applicable to SAD. Gender classification can also be helpful for social robots as part of the smart home environment [33]. The availability of extended speech databases also enables a combined approach, whereby age and gender were processed in parallel [34,35].…”

Speech-based gender recognition using linear prediction and mel-frequency cepstral coefficients

“…Another field of importance that requires gender knowledge is emotion recognition for human-robot interaction since it is one of the most influential factors. Gender recognizer was incorporated in [15] and achieved accuracy closest to 98% to drive speech emotion model based on the first stage detection so that the system accuracy could be improved. Although recent research has proposed the use of deep learning in gender classification to be robust and accurate, a limitation of this type of algorithm is its resource hungry [12], [16], [17].…”