Browsing by Author "Mesgarani, Nima"
Now showing 1 - 2 of 2
Results Per Page
Sort Options
Item Discrimination of Speech From Non-Speech Based on Multiscale Spectro-Temporal Modulations(2005-05-16) Mesgarani, Nima; Shamma, Shihab; Electrical Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)We describe a content-based audio classification algorithm based on novel multiscale spectrotemporal modulation features inspired by a model of auditory cortical processing. The task explored is to discriminate speech from non-speech consisting of animal vocalizations, music and environmental sounds. Although this is a relatively easy task for humans, it is still difficult to automate well, especially in noisy and reverberant environments. The auditory model captures basic processes occurring from the early cochlear stages to the central cortical areas. The model generates a multidimensional spectro-temporal representation of the sound, which is then analyzed by a multi-linear dimensionality reduction technique and classified by a Support Vector Machine (SVM). Generalization of the system to signals in high level of additive noise and reverberation is evaluated and compared to two existing approaches [1] [2]. The results demonstrate the advantages of the auditory model over the other two systems, especially at low SNRs and high reverberation.Item Representation of speech in the primary auditory cortex and its implications for robust speech processing(2008-08-05) Mesgarani, Nima; Shamma, Shihab; Electrical Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Speech has evolved as a primary form of communication between humans. This most used means of communication has been the subject of intense study for years, but there is still a lot that we do not know about it. It is an oft repeated fact, that even the performance of the best speech processing algorithms still lags far behind that of the average human, It seems inescapable that unless we know more about the way the brain performs this task, our machines can not go much further. This thesis focuses on the question of speech representation in the brain, both from a physiological and technological perspective. We explore the representation of speech through the encoding of its smallest elements - phonemic features - in the primary auditory cortex. We report on how population of neurons with diverse tuning properties respond discriminately to phonemes resulting in explicit encoding of their parameters. Next, we show that this sparse encoding of the phonemic features is a simple consequence of the linear spectro-temporal properties of the auditory cortical neurons and that a Spectro-Temporal receptive field model can predict similar patterns of activation. This is an important step toward the realization of systems that operate based on the same principles as the cortex. Using an inverse method of reconstruction, we shall also explore the extent to which phonemic features are preserved in the cortical representation of noisy speech. The results suggest that the cortical responses are more robust to noise and that the important features of phonemes are preserved in the cortical representation even in noise. Finally, we explain how a model of this cortical representation can be used for speech processing and enhancement applications to improve their robustness and performance.