International Conference on NONLINEAR SPEECH PROCESSING,  NOLISP 2015

Jointly organized with the  25th Italian Workshop on Neural Networks, WIRN 2015 

NEWS: Submission deadline exteded to April , 2015. See Dates.

The Non-Linear Speech Processing (NOLISP) workshop is a biennial international workshop aiming at presenting and discussing new ideas, techniques and results related to alternative approaches in speech processing. New and innovative approaches and their applications are welcome to participate in this workshop.

All fields of speech processing are targeted by the workshop, namely:

  • Speech Production
  • Speech Analysis and Modeling
  • Speech Coding
  • Speech Synthesis
  • Speech Recognition
  • Speaker Identification / Verification
  • Speech Enhancement / Separation
  • Speech Perception
  • Others

Invited speakers

  • Jonathan Harringthon (

           Title of the talk: Contextual ambiguities in speech signals and their consequences for sound change

    Abstract: A task that is central to human and machine processing is how the infinite variation in speech signals is related to categories like consonants and vowels that function to distinguish meaning. Modelling historical sound change can shed new light on this issue because of the need to understand how variation over a period of time can bring aboutpermanent category changes. The starting point for these considerations is that speech production is an adaptation to multiple contexts - the phonetic environment, the social context of the speaking situation, the interlocutor to mention but a few. Compatibly, an adult listener hears the same acoustic speech signals differently depending on the contexts in which it was believed to be produced. Sound change according to some influential models may arise because listeners occasionally decontextualize speech by normalising insufficiently or inappropriately for context. One of the issues to be considered is whether this happens because the transmission of dynamic information in speech is inherently ambiguous for certain sequences of sounds and whether such ambiguities are exacerbated in a more casual, less precise speaking style. Another is to do with variation between listeners: that is, sound change may arise because of listener differences in how speech dynamics are processed. A final task will be to assess the extent to which the spread of sound change through the community is brought about by the spontaneous and often subtle ways that interlocutors imitate each other in dialogue.

  • Alessandro Vinciarelli (

    Title of the talk: The social life of features extracted from speech (and from other interesting behaviours).

    Abstract: Social psychologists show that everything observable we do influences what others think about us. Such a phenomenon is so natural and spontaneous that it takes place not only when others meet us face-to-face, but also when the communication is technology mediated (e.g., via phone). The presentation will show that low-level features extracted from speech covariate with the attribution of personality traits and the perception of major social phenomena such as conflict and role. Furthermore, it will show that this can be used to build socially intelligent machines, i.e. technologies capable of dealing with humans like humans do. In a similar vein, the talk will show that similar effects can be observed for other common behaviours in everyday life.