CirdoX: an on/off-line multisource speech and sound analysis software

CirdoX: an On/Off-line Multisource Speech and Sound Analysis Software F. Aman, M. Vacher, F. Portet, W. Duclot, B. Lecouteux CNRS, LIG, F-38000 Grenoble, France Univ. Grenoble Alpes, LIG, F-38000, Grenoble, France Frédéric.Aman@imag.fr, Michel.Vacher@imag.fr, Francois.Portet@imag.fr Abstract Vocal User Interfaces in domestic environments recently gained interest in the speech processing community. This interest is due to the opportunity of using it in the framework of Ambient Assisted Living both for home automation (vocal command) and for call for help in case of distress situations, i.e. after a fall. CIRDOX, which is a modular software, is able to analyse online the audio environment in a home, to extract the uttered sentences and then to process them thanks to an ASR module. Moreover, this system perfoms non-speech audio event classification; in this case, specific models must be trained. The software is designed to be modular and to process on-line the audio multichannel stream. Some exemples of studies in which CIRDOX was involved are described. They were operated in real environment, namely a Living lab environment. Keywords: audio and speech processing, natural language and multimodal interactions, Ambient Assisted Living (AAL). 1. Introduction Voice User Interfaces (VUI) are becoming more and more popular in ambient intelligence environments whether it is for smartphones or for smart homes (e.g., S.A.H.R.A.1). For instance, in domestic environments, VUI recently gained interest in the speech processing community as ex- emplified by the rising number of smart home projects that consider Automatic Speech Recognition (ASR) in their de- sign (Istrate et al., 2008; Badii and Boudy, 2009; Hamill et al., 2009; Filho and Moir, 2010; Gemmeke et al., 2013; Christensen et al., 2013; Cristoforetti et al., 2014; Vacher et al., 2015b). However, the speech technology, though mature is still dif- ficult to master for non-expert and to tune for a particular usage (Vacher et al., 2015a). Although many good qual- ity commercial systems were made available during this decade, these are neither open source nor customisable by the user. Moreover, for some of them, respect to user privacy may be questioned. Hence, it becomes important for research in natural language interaction to access open- source and highly configurable real-time VUI software. Many toolboxes and libraries exist for audio processing such as Planet CCRMA at Home (CCRMA, 2016), ChucK (ChucK, 2016), etc. But these are dedicated for music pro- cessing and not for speech interaction. General purpose recorder and editor such as SoX (Sox, 2016) and Audacity (Audacity, 2016) make it possible to acquire audio signals in real-time and edit them, but these tools are mainly de- signed to perform off-line processing. Regarding speech interaction, there are a high number of ASR (Automatic Speech Recognition) systems2 but again, using them in an on-line hand free setting requires a high expertise in signal and speech processing. Ideally, a complete hand free VUI system would be com- posed of five main components: 1. an multisource acquisition stage; 2. a signal enhancement stage; 1http://sarah.encausse.net 2https://en.wikipedia.org/wiki/List_of_ speech_recognition_software 3. a Voice Activity Detection (VAD) stage; 4. an Automatic …

Content truncated to 500 words