Modern smartphones present an ideal device for pervasive sensing of human behaviour. Microphones have the potential to reveal key information about a persons behaviour.However, they have been utilized to a significantly lesser extent than other smartphone sensors in the context of human behaviour sensing. We postulate that, in order for microphones to be useful in behaviour sensing applications, the analysis tecniques must be flexible and allow easy modification of the types of sounds to be sensed. A simplification of the training data collection process could allow a more flexible sound classification framework. We hypothesize that detailed training, a prerequisite for the majority of sound sensing techniques, is not necessary and that a significantly less detailed and time consuming data collection process can be carried out, allow-ng even a non expert to conduct the collection, labeling, and training process. To test this hypothesis, we implement a diverse density-based multiple instance learning framework, to identify a target sound, and a bag trimming algorithm, which, using the target sound, automatically segments weakly labeled soundclips to construct an accurate training set. Experiments reveal that our hypothesis is a valid one and results show that classifiers, trained using the automatically segmented training sets,were able to accurately classify unseen sound samples with accuracies comparable to supervised classifiers, achieving an average F-measure of 0.969 and 0.87 for two weakly supervised datasets.
展开▼
