Impulse response shaping (IRS) is a prefiltering technique for modifying the characteristics of a linear channel to achieve desirable characteristics. Acoustic impulse response shaping is used to apply partial equalization to reduce the effects of reverberation on audio signals propagating inside a room and is thus used for listening room compensation. To be capable of responding rapidly to time varying room acoustics, a shaping algorithm must be computationally fast; this is difficult to achieve since the filters to be designed may have thousands of coefficients. This paper derives variants of the Dual Augmented Lagrangian Method, presents their application to the design of the shaping filters, and validates their efficacy through experimental studies. In addition, the formulations have been modified to design shaping filters that are robust to change in microphone positions.

Spatial audio may be performed in rooms using active-compensated sound field reproduction (AC-SFR), to cancel the reverberant reflections. Reproduction of a sound field may be performed using precalculated loudspeaker filters. However the robustness can be poor due to the temperature-induced perturbation of the room impulse responses (RIR). To compensate for reverberation, filters with significant non-causal components are designed. It is known that the requirements on the loudspeaker filters for room equalization may be reduced by applying the method of impulse response shaping (IRS). In this paper, we show that actively compensating paths with a long propagation distance are more heavily influenced by temperature than short paths. We hence impose a shaping penalty on the non-causal components of designed loudspeaker filters, to reduce the error caused by a change in temperature. We obtain AC-SFR designs that are more robust to changes in temperature.

Neural task-oriented dialogue systems often struggle to smoothly interface with a knowledge base.In this work, we seek to address this problem by proposing a new neural dialogue agent that is able to effectively sustain grounded, multi-domain discourse through a novel key-value retrieval mechanism.The model is end-to-end differentiable and does not need to explicitly model dialogue state or belief trackers.We also release a new dataset of 3,031 dialogues that are grounded through underlying knowledge bases and span three distinct tasks in the in-car personal assistant space: calendar scheduling, weather information retrieval, and point-of-interest navigation.Our architecture is simultaneously trained on data from all domains and significantly outperforms a competitive rulebased system and other existing neural dialogue architectures on the provided domains according to both automatic and human evaluation metrics.

Acoustic impulse response shaping is a prefiltering technique to reduce the perceptible effects of room reverberation. The room impulse responses to be shaped must first be measured, and these measurements can contain errors. Furthermore, room responses vary with changes in temperature and humidity and also with changes in measurement position. This letter presents a method for enhancing the robustness of the shaping procedure so that the effects of these errors and changes are minimized. The method uses a stochastic model of the channel variations to explicitly limit the probability of large deviations from the desired performance. The method is evaluated on realistic channel perturbations and the resulting shaped responses are shown to comply with the robustness specification.

Impulse response shaping is a technique for partly equalizing impulse responses. In acoustics, it can be used for the reproduction of audio signals mitigated by distortions in a room. The most significant phenomenon among the distortions is reverberation, a straightforward characterization of which is the room impulse response. Room responses can be characterized but could contain measurement errors or noise. In addition, room responses vary with changes in atmospheric conditions such as temperature and humidity and also due to change in positions inside a room. The design of a shaping filter robust to at least some of these variations is likely to be very useful, which is considered in this work. The method uses a computationally efficient approach based on Basis Pursuit DeNoising (BPDN).

A new approach for the segregation of monaural sound mixtures is presented based on the principle of temporal coherence and using auditory cortical representations. Temporal coherence is the notion that perceived sources emit coherently modulated features that evoke highly-coincident neural response patterns. By clustering the feature channels with coincident responses and reconstructing their input, one may segregate the underlying source from the simultaneously interfering signals that are uncorrelated with it. The proposed algorithm requires no prior information or training on the sources. It can, however, gracefully incorporate cognitive functions and influences such as memories of a target source or attention to a specific set of its attributes so as to segregate it from its background. Aside from its unusual structure and computational innovations, the proposed model provides testable hypotheses of the physiological mechanisms of this ubiquitous and remarkable perceptual ability, and of its psychophysical manifestations in navigating complex sensory environments.