Aiming at detecting pronunciation errors produced by second language learners and providing corrective feedbacks related with articulation, we address effective articulatory models based on deep neural network (DNN). Articulatory attributes are defined for manner and place of articulation. In order to efficiently train these models of non-native speech without such data, which is difficult to collect in a large scale, several transfer learning based modeling methods are explored. We first investigate three closely-related secondary tasks which aim at effective learning of DNN articulatory models. We also propose to exploit large speech corpora of native and target language to model inter-language phenomena. This kind of transfer learning can provide a better feature representation of non-native speech. Related task transfer and language transfer learning are further combined on the network level. Compared with the conventional DNN which is used as the baseline, all proposed methods improved the performance. In the native attribute recognition task, the network-level combination method reduced the recognition error rate by more than 10% relative for all articulatory attributes. The method was also applied to pronunciation error detection in Mandarin Chinese pronunciation learning by Japanese native speakers, and achieved the relative improvement up to 17.0% for detection accuracy and up to 19.9% for F-score, which is also better than the lattice-based combination.

Let Gs=(Vs, Es) be a simple connected graph. A vertex v ∈ Vs is an articulation vertex if deletion of v and its incident edges from Gs disconnects the graph into at least two connected components. Finding all articulation vertices of a given graph is called the articulation vertex problem. A vertex u ∈ Vs is called a hinge vertex if there exist any two vertices x and y in Gs whose distance increase when u is removed. Finding all hinge vertices of a given graph is called the hinge vertex problem. These problems can be applied to improve the stability and robustness of communication network systems. In this paper, we propose linear time algorithms for the articulation vertex problem and the hinge vertex problem of circular permutation graphs.

When a micromachine works inside a narrow space inside tubes and equipment such as a microfactory, a microdevice that has a visual function is indispensable. To monitor the minute shapes of microfabrication and microassembly process that are impossible to observe, fiber-optic sensors and actuators for environmental recognition devices have been developed. The devices are designed to allow stereoscopic and microscopic observation and to measure the dimensions of microparts. To achieve these goals and to realize minute structures and functions, we developed environmental recognition devices for microfabrication process with functions of far and near field observation, tactile sensing and tip articulation, for microassembly process with functions of stereoscopic observation and tip articulation. The results show that easy and safe environmental recognition is possible in the narrow spaces of a microfactory.

Cerebellar Model Articulation Controller (CMAC) has been proposed as one of artificial neural networks. This paper describes a design scheme of intelligent control system consists of some CMACs. Each of CMACs is trained for the specified reference signal. A new CMAC is generated for unspecified reference signals, and the CMAC whose reference signal is nearest for the new reference signal, is eliminated. Therefore, since the reference signals are removed from the input signals of the CMAC, the proposed intelligent controller can be designed with fairly small memories.

The mesh-connected computers with hyperbus broadcasting are an extension of the mesh-connected computers with multiple broadcasting. Instead of using local buses, we use global buses to connect processors. Such a strategy efficiently reduces the time complexity of the semigroup problem from O(N) to O(log N). Also, the matrix multiplication and the transitive closure problems are solved in O(log N) and O(log2 N) time, respectively. Then, based on these operations, several interesting problems such as the connected recognition problem, the articulation problem, the dominator problem, the bridge problem, the sorting problem, the minimum spanning tree problem and the bipartite graph recognition problem can be solved in the order of polylogarithmic time.

This paper gives a model to explain one phenomenon found in the process of creative concept formation, i.e. the phenomenon that people often get trapped in some state where the mental world remains nebulous and sometimes suddenly make a jump to a new concept. This phenomenon has been qualitatively explained mainly by the philosophers but there have not been models for explaining it quantitatively. Such model is necessary in a new research field to study the systems for aiding human creative activities. So far, the work on creation aid has not had theoretical background and the systems have been built based only on trial and error. The model given in this paper explains some aspects of the phenomena found in creative activities and give some suggestions for the future systems for aiding creative concept formation.

Binaural effects in two measures are studied. One measure is the detectable limen of click sounds under lateralization of diotic or dichotic noise signals, and the other is phoneme articulation score under localization or lateralization of speech and noise signals. The experiments use a headphones system with listener's own head related transfer function (HRTF) filters. The HRTF filter coefficients are calculated individually from the impulse responses due to the listener's HRTF measured in a slightly sound reflective booth. The frequency response of the headphone is compensated for using an inverse filter calculated from the response at the subject's own ear canal entrance point. Considering the speech frequency band in tele-communication systems is not sufficiently wide, the bandwidth of the HRTF filter is limited below 6.2 kHz. However, the experiments of the localization simulation in the horizontal plane show that the sound image is mostly perceived outside the head in the simulated direction. Under simulation of localization or lateralization of speech and noise signals, the phoneme articulation score increases when the simulation spatially separates the phonemes from the noise signals while the total signal to noise ratio for both ears is maintained constant. This result shows the binaural effect in speech intelligibility under the noise disturbance condition, which is regarded as a part of the cocktail party effect.