Speech Enhancement With Inventory Style Speech Resynthesis

We present a new method for the enhancement of speech. The method is designed for scenarios in which targeted speaker enrollment as well as system training within the typical noise environment are feasible. The proposed procedure is fundamentally different from most conventional and state-of-the-art denoising approaches. Instead of filtering a distorted signal we are resynthesizing a new “clean” signal based on its likely characteristics. These characteristics are estimated from the distorted signal. A successful implementation of the proposed method is presented. Experiments were performed in a scenario with roughly one hour of clean speech training data. Our results show that the proposed method compares very favorably to other state-of-the-art systems in both objective and subjective speech quality assessments. Potential applications for the proposed method include jet cockpit communication systems and offline methods for the restoration of audio recordings.

Effects of training and melodic features on mode perception

The two modes most widely used in Western music today convey opposite moods—a distinction that nonmusicians and even young children are able to make. However, the current studies provide evidence that, despite a strong link between mode and affect, mode perception is problematic. Nonmusicians found mode discrimination to be harder than discrimination of other melodic features, and they were not able to accurately classify major and minor melodies with these labels. Although nonmusicians were able to classify major and minor melodies using affective labels, they performed at chance in mode discrimination. Training, in the form of short lessons given to nonmusicians and the natural musical experience of musicians, improved performance, but not to ceiling levels. Tunes with high note density were classified as major, and tunes with low note density as minor, even though these features were actually unrelated in the experimental material. Although these findings provide support for the importance of mode in the perception of emotion, they clearly indicate that these mode perceptions are inaccurate, even in trained individuals, without the assistance of affective labeling.

Musical Expertise and Melodic Structure in Memory for Musical Notation

Two experiments plus a pilot investigated the role of melodic structure on short-term memory for musical notation by musicians and nonmusicians. In the pilot experiment, visually similar melodies that had been rated as either "good" or "bad" were presented briefly, followed by a 15-sec retention interval and then recall. Musicians remembered good melodies better than they remembered bad ones: nonmusicians did not distinguish between them. In the second experiment, good, bad, and random melodies were briefly presented, followed by immediate recall. The advantage of musicians over nonmusicians decreased as the melody type progressed from good to bad to random. In the third experiment, musicians and nonmusicians divided the stimulus melodies into groups. For each melody, the consistency of grouping was correlated with memory performance in the first two experiments. Evidence was found for use of musical groupings by musicians and for use of a simple visual strategy by nonmusicians. The nature of these musical groupings and how they may be learned are considered. The relation of this work to other studies of comprehension of symbolic diagrams is also discussed.