Akitoshi Itai

Bio: Akitoshi Itai is an academic researcher from Chubu University. The author has contributed to research in topics: Background noise & Extremely low frequency. The author has an hindex of 8, co-authored 49 publications receiving 202 citations. Previous affiliations of Akitoshi Itai include Aichi Prefectural University.

Footstep classification using simple speech recognition technique

Abstract: The characteristics of human footsteps are determined by the gait, the footwear and the floor. Accurate footstep analysis would be useful in various applications, home security service, surveillance and understanding of human action since the gait expresses personality, age and gender. The feasibility of a footstep classification has been confirmed by using the acoustic feature parameter[1], however, almost of conventional approaches are focused on the statistical features and pattern recognitions. In the speech recognition, a feature string is stretched and compressed in the time domain. The dynamic programming is used to accomplish this task, and is an effective method of absorbing time domain fluctuations. In the footstep classification, footstep sound (i.e. an impact sound and a fricative sound) is expanded and contracted in the time domain the same as speeches. This paper applies the DTW and cepstra to the footstep classification. Result shows that the proposed method is useful to the footstep recognition problems.

Footstep Recognition with Psyco-acoustics Parameter

Abstract: The characteristics of a footstep are determined by the gait, the footwear and the floor. Accurate footstep analysis would be useful in various applications, home security service, surveillance and understanding of human action since the gait expresses personality, age and gender. The feasibility of personal identification has been confirmed by using the feature parameter of footsteps (Shoji et al., 2004), however, the recognition rate of this method decreases as the number of subjects increases. This paper applies psycho-acoustics parameter to feature extraction. Results show that the parameter proposed herein yields effective and practical personal identification

Personal Identification Using Footstep Detection in In-Door Environment

Abstract: Footsteps, with different shoes of heels, sneakers, leathers or even bare footed, will appear in different grounds of concrete, wood, etc. If a footstep is discriminable, the application to various fields can be considered. In this paper, the feature extraction of a footstep is investigated. We focus on the recognizing a certain kind of footstep waveforms under the restricted condition. We propose a new methodology using the feature parameter such as the peak frequency set by the mel-cepstrum analysis, the walking intervals and the similarity of spectrum envelope. It is shown for personal identification that the performance of the proposed method is effective.

Sound Localization of Approaching Vehicles Using Uniform Microphone Array

Abstract: Due to recent trends in society, enhanced car safety is strongly required. As one piece of environmental traffic information, the direction of approaching vehicles can be estimated from an analysis of sound sources. This gives information of approaching vehicles to drivers at intersections when visibility is poor. To prevent traffic accidents, it is necessary to detect the direction of approaching vehicles more accurately and rapidly. This paper describes a scheme that uses a uniform linear microphone array with four microphones to estimate the direction of approaching vehicles. We employ a correlation method to realize sound source localization and we propose a direction estimation method with cubic spline interpolation in Carter, G. C. and Abraham, P. B. (1980). Trials show that the proposed scheme offers good performance.

Global Noise Estimation Based on Tensor Product Expansion with Absolute Error

Abstract: This paper proposes a novel signal estimation method that uses a tensor product expansion. When a bivariable function, which is expressed by two-dimensional matrix, is subjected to conventional tensor product expansion, two single variable functions are calculated by minimizing the mean square error between the input vector and its outer product. A tensor product expansion is useful for feature extraction and signal compression, however, it is difficult to separate global noise from other signals. This paper shows that global noise, which is observed in almost all input signals, can be estimated by using a tensor product expansion where absolute error is used as the error function.

Pattern Recognition and Machine Learning

Abstract: Probability Distributions.- Linear Models for Regression.- Linear Models for Classification.- Neural Networks.- Kernel Methods.- Sparse Kernel Machines.- Graphical Models.- Mixture Models and EM.- Approximate Inference.- Sampling Methods.- Continuous Latent Variables.- Sequential Data.- Combining Models.

Accelerometer-Based Gait Recognition by Sparse Representation of Signature Points With Clusters

Abstract: Gait, as a promising biometric for recognizing human identities, can be nonintrusively captured as a series of acceleration signals using wearable or portable smart devices. It can be used for access control. Most existing methods on accelerometer-based gait recognition require explicit step-cycle detection, suffering from cycle detection failures and intercycle phase misalignment. We propose a novel algorithm that avoids both the above two problems. It makes use of a type of salient points termed signature points (SPs), and has three components: 1) a multiscale SP extraction method, including the localization and SP descriptors; 2) a sparse representation scheme for encoding newly emerged SPs with known ones in terms of their descriptors, where the phase propinquity of the SPs in a cluster is leveraged to ensure the physical meaningfulness of the codes; and 3) a classifier for the sparse-code collections associated with the SPs of a series. Experimental results on our publicly available dataset of 175 subjects showed that our algorithm outperformed existing methods, even if the step cycles were perfectly detected for them. When the accelerometers at five different body locations were used together, it achieved the rank-1 accuracy of 95.8% for identification, and the equal error rate of 2.2% for verification.

Comparative Analysis and Fusion of Spatiotemporal Information for Footstep Recognition

Abstract: Footstep recognition is a relatively new biometric which aims to discriminate people using walking characteristics extracted from floor-based sensors. This paper reports for the first time a comparative assessment of the spatiotemporal information contained in the footstep signals for person recognition. Experiments are carried out on the largest footstep database collected to date, with almost 20,000 valid footstep signals and more than 120 people. Results show very similar performance for both spatial and temporal approaches (5 to 15 percent EER depending on the experimental setup), and a significant improvement is achieved for their fusion (2.5 to 10 percent EER). The assessment protocol is focused on the influence of the quantity of data used in the reference models, which serves to simulate conditions of different potential applications such as smart homes or security access scenarios.