Consumer electronics equipped with a microphone array, such as car navigation devices and headsets commonly implement speech enhancement techniques based on the gradient method to cope additive noise. However, while these had been originally developed for voice communication can maximize signal-to-distortion ratio (SDR), they cannot always automatic recognition (ASR) accuracy. For this reason, front-end parameters have adjusted by human experts each environment acoustic model. In study, we...

We propose a method for estimating the prior signal-to-noise ratio (SNR), which is used calculating Wiener filter distant sound source extraction, from output signals of beamforming using statistical mapping based on deep neural network (DNN). Since informative features to estimate SNR are included in multiple outputs, can be accurately estimated by this DNN. The proposed was applied large microphone array, design optimized form effective directivity patterns extract sources. Experimental...

Microphone array Wiener filtering has been studied for practical sound source enhancement in noisy environments. In our previous work, the power spectral density (PSD)-estimation-in-beamspace method was proposed estimating PSDs of target sources and surrounding noises. Assuming that signals are uncorrelated with each other, relationships among multiple beamforming outputs (beamforming PSDs), those (source beam patterns can be modelled by a simple linear mixture domain. To estimate PSDs,...

Under acute stress conditions in the rat, there is rapid and transient increase circulating prolactin (PRL). This leads to an elevated expression of long form PRLR (PRLR-L) first hypothalamus choroid plexus. PRL involved inhibition stress-induced hypocalcemia gastric erosion. In this study we used rat a morpholino-antisense oligonucleotide elucidate mechanism by which hypothalamic mediates restraint water (RSW)-induced We found that effect largely mediated PRLRs paraventricular nucleus...

The Wiener filter has been used as a post-filter applied to the output of beamforming, which boosts overall performance sound source enhancement. Since power spectral density (PSD) each needs be estimated derive filter, previous study attempted estimate PSDs from signals multiple beamformings using linear approximation realized by least squares method. In this study, we propose an alternative approach estimation process that uses neural network implement non-linear function. Experimental...

A method for constructing deep neural networks (DNNs) accurate supervised source enhancement is proposed. Attempts were made in previous studies to estimate the power spectral densities (PSDs) of sound sources, which are used Wiener filters enhancement, from output multiple beamformings using DNNs. Although performance improved, it was not possible guarantee PSD estimation since trained DNNs treated as black boxes. The proposed DNN construction uses non-negative auto-encoders and...

We have developed a self-adjustable speech enhancement and recognition (SSER) system to make automatic (ASR) robust variation of acoustic conditions. The SSER enhances components in observed signals, switching parameter values accordance with the Candidates for value are automatically generated using real-coded genetic algorithm (GA) advance. experimental results show that performance parameter-set is improved by proposed method.

We propose a technique of multi-channel speech enhancement based on integration beamforming and statistical model-based to clearly extract the target speech, even in very noisy environments. Conventional microphone array-based techniques estimate noise power spectral densities (PSDs) from spatial cues sound sources; however, their estimation errors dramatically increase when there are many sources. integrated clean models trained advance PSDs estimated beamspace compose observation designed...

Noise reduction has been extensively studied to prevent the accuracy of automatic speech recognition (ASR) being severely degraded due various noise sources observed in practical use. The effect is usually dependent on parameters used noise-reduction method, which need be tuned for different types achieve best accuracy. We propose a method automatically switching such way that ASR maximized. proposed assigns most suitable parameter set each sentence by measuring and grouping characteristics...

Tendon-driven systems are one of efficient methods to downsize robots. This paper discusses application tension control into a joint using cable differential. Cable differential is mechanism which allows two-degrees freedom (DOF) in joint. Controllers for the differential-driven proposed paper. The actuated by four linear motors. controllers designed regarding tendon-driven system as redundant system. In this way, considered task null space. makes implementation stereotypical.

We propose a noise-suppressing microphone array using power spectrum density (PSD) estimation. The microphones of the proposed achieve over 22 dB higher signal-to-noise ratio on average than other commercial microphones. This is useful for voice communication and recognition in highly noisy environments such as factories construction sites.

We propose a microphone array speech enhancement method that integrates spatial-cue-based source power spectral density (PSD) estimation and statistical model-based PSD estimation. The goal of this research was to clearly pick up target even in noisy environments such as crowded places, factories, cars running at high speed. Beamforming with post-Wiener filtering is commonly used many conventional studies on microphone-array noise reduction. For calculating Wiener filter, speech/noise PSDs...