- Speech and Audio Processing
- Speech Recognition and Synthesis
- User Authentication and Security Systems
- Advanced Adaptive Filtering Techniques
- Internet Traffic Analysis and Secure E-voting
- Acoustic Wave Phenomena Research
- Music and Audio Processing
- Cryptography and Data Security
- Gaze Tracking and Assistive Technology
- Advanced Data Compression Techniques
- Context-Aware Activity Recognition Systems
- Hand Gesture Recognition Systems
- Speech and dialogue systems
- Education and Learning Interventions
- Tactile and Sensory Interactions
- Robotics and Automated Systems
- Neural Networks and Applications
- Human auditory perception and evaluation
- Caching and Content Delivery
- IPv6, Mobility, Handover, Networks, Security
- Access Control and Trust
- Wireless Body Area Networks
- Antenna Design and Optimization
- Network Security and Intrusion Detection
- Advanced Steganography and Watermarking Techniques
Sojo University
2020-2024
Kanagawa Institute of Technology
2010-2023
NTT (Japan)
2003-2020
Obihiro University of Agriculture and Veterinary Medicine
2013
Waseda University
2007-2010
Toshiba (Japan)
2005
Kyoto University
2003
Interface (United Kingdom)
1996-1999
We have developed a method of segregating desired speech from concurrent sounds received by two microphones. In this method, which we call SAFIA, signals microphones are analyzed discrete Fourier transformation. For each frequency component, differences in the amplitude and phase between channels calculated. These used to select components signal that come direction reconstruct these as source signal. To clarify effect resolution on proposed conducted three experiments. First, relationship...
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...
A stereo microphone array developed for a high-definition videophone system is presented. The consists of pair fixed beamformers to collect sounds in clearly while suppressing the far-end talk emitted from loudspeakers decrease undesirable influence acoustic echo. Both objective and subjective experimental results demonstrate that satisfies required specifications, which have not been achieved by several conventional schemes.
A password-based authentication is a simple technique used for user of computer or Web services. However, by shoulder surfing passwords can be stolen easily. To prevent these attacks, we need to shield our keypad from view using body cupping hand. hide them easily, it better that itself the range inputting keys small. We propose password system with only three and show an analysis security evaluation results short test. Our method also utilized handicapped users who find hard use input in...
We have developed a sound image rendering system for headphones that enables anyone to easily create realistic 3-D sound. This paper presents the structure of this system, and results subjective tests confirm system's feasibility in affecting has been used making music CDs.
This paper presents the fragility function for wooden house obtained from field investigation of flood disaster in Nakanoshima town, Niigata prefecture due to July 13, 2004 heavy rainfall. At emergency phase disasters, early understanding overview damage is required management personnel and resource. To understand easily, which estimates inundation depth helpful. The measured at many points rank each identified this disaster. constructed using these data. Additionally, we marshaled those...
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.
Abstract The distribution of secret information so that no one knows which is being passed to whom, such as in a poker game or the like, an important issue for security. Secret expected have applications key distribution, electronic voting, and on. A method must maintain secrecy untraceability distributed content, without requiring any effort on part recipient. In this paper, authors propose method, using third‐party organizations, recipient can easily receive having engage work itself; be...
We propose a method for achieving sharp directivity by sensing signals in diffuse acoustic field. Directivity control based on beamforming has been studied to make it possible extract the waveform and location of an identified target source even if there are many noise sources. Sharp can be achieved minimizing output power filter. However, is difficult minimize over broad frequency ranges. Our approach spatial properties transfer functions correlation matrix, using reflector that surrounds...
In this paper, we propose a novel training strategy for deep neural network (DNN) based small-footprint acoustic models. The accuracy of DNN-based automatic speech recognition (ASR) systems can be greatly improved by leveraging large amounts data to improve the level expression. DNNs use many parameters enhance performance. Unfortunately, resource-constrained local devices are unable run complex ASR systems. For building compact models, knowledge distillation (KD) approach is often used. KD...
Several studies have reported that experts outperform novices in specific domains. However, the superiority of accuracy, taking both trueness and precision into consideration, has not yet been explored. Here, we examined differences between expert novice performances by evaluating accuracy their estimations physical concentrations sodium chloride solutions while employing a visual analog scale. In Experiment 1, 14 13 tasted 6 until they had learned intensities. Subsequently, repeatedly rated...
The vast majority of online servicesrequire some form ofpersonal authentication.Unfortunately, standard password authenticationstrikes a poor balance between security and convenience, whereas strongerauthentication schemes,such as those involvingbiometrics, one-time passwords, electronic certificates, depend on specialized hardware and/orhardware tokens.To achieve robustness, cost-effectiveness together, we propose scheme for dynamicknowledge-based authentication in which Twitter direct...
In this paper, we propose a personal portable proxy server that offers various functions from USB flash drive.With the rapidly growing volume of data generated by applications, managing becomes problem.A drive can help us manage easily, because carry any and application anytime on drive.Therefore, an intelligent with drive.This provides convenient features, such as content filtering, history sharing, phishing prevention, especially automatic form completion.We also using to exchange...
When an ultrasound vibration that has been amplitude-modulated by audible sound is transmitted to auricular cartilage such as the tragus, heard. To study this phenomenon, we measured nonlinear effect using a piece of pig's and acceleration sensor, found self-demodulation occurred in cartilage. We also pressure level acoustic coupler simulated our earphone placed on tragus. The results show was generated coupler's air chamber. However, most important factor from produced Therefore, think perceived
When finite-state machines are used for dialogue models of a conversational agent, learning algorithms which learn probabilistic automata with the state merging method useful. However these should whole data every time number example dialogues increases. Therefore, cost is large when we construct gradually. We proposed decreases compatibility checks by caching information, and evaluated it perplexities learned models. From comparison among models, caches only compatibility-changed states...
A stereo microphone array developed for a high-definition videophone system is presented. It consists of pair fixed beamformers to collect sounds in clearly while suppressing the far-end talk decrease undesirable influence acoustic echo. Experimental results prove that satisfies required specifications, which have not been achieved with conventional schemes.
In teleconferencing, conversation should flow naturally between separate sites. The acoustic design should, of course, be able to at least convey the correct volume. However, sound localization is also very important for realizing kind virtual reality world that teleconferencing. Coordinating visual image (a speaker’s face) and (his or her voice) creates a environment in which participants each site feel as if they are having same room. Two prototype teleconferencing systems with effects...
The acquisition of comprehensive sound field information is a central topic in spatial acoustics. In conventional systems, recording devices must be located at the listener’s viewpoint. This research introduces technique for spherical microphone arrays which makes use a-priori regarding distribution sources. proposal generates descriptions viewpoints that are away from device, as long there no sources between target viewpoint and array. Sound descriptions, set harmonic expansion...
Mixing information is a key technique in network security. In mental poker, Internet lottery, and Mix-Net for electronic voting, we use shuffling to maintain anonymity, privacy, fairness. this paper, propose method confirm whether mix done well by finding fixed point the system while keeping how shuffle was secret.