We propose a new algorithm for blind source separation (BSS), in which independent component analysis (ICA) and beamforming are combined to resolve the slow-convergence problem through optimization ICA. The proposed method consists of following three parts: (a) frequency-domain ICA with direction-of-arrival (DOA) estimation, (b) null based on estimated DOA, (c) integration diversity both iteration frequency domain. unmixing matrix obtained by is temporally substituted iterative optimization,...

The Takemaru-kun system is a real world speech-oriented guidance located at the Ikoma-City North Community Center. has been operated daily from November, 2002, to provide visitors speech interface for information retrieval. This also aims field test of and collecting actual utterance data. By analyzing evaluating collected utterances, flexible processing requirements are discovered according user's age group. It becomes impossible disregard increase child users when installed in public...

We address a method to efficiently select Gaussian mixtures for fast acoustic likelihood computation. It makes use of context-independent models selection and back-off corresponding triphone models. Specifically, the k-best phone by preliminary evaluation, higher resolution are applied, others assigned likelihoods with monophone This scheme assigns more reliable un-selected states than conventional based on VQ codebook. can also incorporate efficient pruning at which offsets increased size...

We implemented a humanoid robot, ASKA, in our university reception desk for the computerized guidance. ASKA can recognize user's question utterance, and answer by its text-to-speech voice, hand gesture head movement. This paper describes speech related parts of ASKA. deal with wide task domain 20k large vocabulary using word trigram model an elaborated speaker-independent acoustic model. also make response keyword key-phrase detection N-best recognition results. The rate is 90.9%,...

Confidence scoring based on word posterior probability is usually performed as a post process of speech recognition decoding, and also needs large number hypotheses to get enough confidence quality. We propose simple way computing the using estimated while decoding. At expansion stack decoding search, local sentence likelihoods that contain heuristic scores unreached segment are directly used compute probabilities. Experimental results showed that, although not optimal, we can provide...

In previous works, we introduced a special device (Non-Audible Murmur (NATM) microphone) able to detect very quietly uttered speech (murmur), which cannot be heard by listeners near the talker. Experimental results showed efficiency of in NAM recognition. Using normal-speech monophone hidden Markov models (HMM) retrained with data from specific speaker, could recognize high accuracy. Although were promising, serious problem is HMM retraining, requires large amount training data. this paper,...

We propose a spatial subtraction array (SSA) and known noise superimposition to achieve noise-robust hands-free speech recognition which can be used in human-robot interaction. In the proposed SSA, reduction is achieved by subtracting estimated power spectrum from target enhanced mel-scale filter bank domain. This offers realization of error-robust spectral with few computational complexities. addition, we introduce technique domain, utilize matched acoustic model for noise. compensate...

Patient and public involvement in research refers to patients or caregivers with disease experience contributing the design, conduct dissemination of results from research. has given rise new fields healthcare-oriented potential transform infectious diseases through interventional trials. Our recommendations best practices years organizing respiratory syncytial virus parent networks are provided.