Content Classification of Multimedia Documents using Partitions of Low-Level Features

Audio-visual documents obtained from German TV news are classified according to the IPTC topic categorization scheme. To this end usual text classification techniques are adapted to speech, video, and non-speech audio. For each of the three modalities word analogues are generated: sequences of syllables for speech, “video words” based on low level color features (color moments, color correlogram and color wavelet), and “audio words” based on low-level spectral features (spectral envelope and spectral flatness) for non-speech audio. Such audio and video words provide a means to represent the different modalities in a uniform way. The frequencies of the word analogues represent audio-visual documents: the standard bag-of-words approach. Support vector machines are used for supervised classification in a 1 vs. n setting. Classification based on speech outperforms all other single modalities. Combining speech with non-speech audio improves classification. Classification is further improved by supplementing speech and non-speech audio with video words. Optimal F-scores range between 62% and 94% corresponding to 50% - 84% above chance. The optimal combination of modalities depends on the category to be recognized. The construction of audio and video words from low-level features provide a good basis for the integration of speech, non-speech audio and video.

Impact Study of Nonverbal Facial Cues on Spontaneous Chatting with Virtual Humans

Non-verbal communication (NVC) is considered to represent more than 90 percent of everyday communication. In virtual world, this important aspect of interaction between virtual humans (VH) is strongly neglected. This paper presents a user-test study to demonstrate the impact of automatically generated graphics-based NVC expression on the dialog quality: first, we wanted to compare impassive and emotion facial expression simulation for impact on the chatting. Second, we wanted to see whether people like chatting within a 3D graphical environment. Our model only proposes facial expressions and head movements induced from spontaneous chatting between VHs. Only subtle facial expressions are being used as nonverbal cues - i.e. related to the emotional model. Motion capture animations related to hand gestures, such as cleaning glasses, were randomly used to make the virtual human lively. After briefly introducing the technical architecture of the 3D-chatting system, we focus on two aspects of chatting through VHs. First, what is the influence of facial expressions that are induced from text dialog? For this purpose, we exploited an emotion engine extracting an emotional content from a text and depicting it into a virtual character developed previously [GAS11]. Second, as our goal was not addressing automatic generation of text, we compared the impact of nonverbal cues in conversation with a chatbot or with a human operator with a wizard of oz approach. Among main results, the within group study -involving 40 subjects- suggests that subtle facial expressions impact significantly not only on the quality of experience but also on dialog understanding.