No abstract

Analysing the human voice has always been a challenge to the engineering society for various purposes such as product review, emotional state detection, developing AI, and much more. Two basic grounds of voice or speech analysis are to detect human gender and the geographical region based on accent. This study presents a three-layer feature extraction method from the raw human voice to detect the gender as male or female, as well as the region from where that voice belongs. Fundamental frequency, spectral entropy, spectral flatness, and mode frequency have been calculated in the first layer of feature extraction. On the other hand, Mel Frequency Cepstral Coefficient has been used to extract the features in the second layer and linear predictive coding in the third layer. Regular voice contains some noises which have been removed with multiple audio data filtering processes to get noise-free smooth data. Multi-Outputbased 1D Convolutional Neural Network has been used to recognize gender and region from a combined dataset which consists of TIMIT, RAVDESS, and BGC datasets. The model has successfully predicted the gender with 93.01% and region with 97.07% accuracy. This method works better than usual state-ofthe-art methods in separate datasets along with the combined dataset on both gender and region classification.

The identification of people's gender and events in our everyday applications by means of gait knowledge is becoming important. Security, safety, entertainment, and billing are the examples of such applications. Many technologies could also be used to monitor people's gender and activities. Existing solutions and applications are subject to the privacy and the implementation costs and the accuracy they have achieved. For instance, CCTV or Kinect sensor technology for people is a violation of privacy, since most people don't want to make their photos or videos during their daily work. A new addition to the gait analysis field is the inertial sensor-based gait dataset. Therefore, in this paper, we have classified people's gender from an inertial sensor-based gait dataset. We have collected the gait dataset from Osaka University. Four machine learning algorithms have been applied to identify people's gender and they are Support Vector Machine (SVM), K-Nearest Neighbor (KNN), Bagging, and Boosting algorithm. For classifying gender, some useful features are extracted from the raw data, and 84 features have been used to identify people's gender. After feature selection the experimental outcome exhibits the accuracy of gender identification via the Bagging stands at around 87.858%, while it is about 86.09% via SVM. This will in turn form the basis to support human wellbeing by using gait knowledge.

No abstract

Identifying people's ages and events by the use of gait information is a popular issue in our daily applications. The most popular application is health, security, entertainment and charging. A variety of algorithms for data mining and deep learning have been proposed. Many different technologies may be used to keep track of people's ages and behaviors. Existing approaches and technologies are limited by their performance, as well as their privacy and deployment costs. For example CCTV or Kinect sensor technology constitutes a privacy offense and most people do not want to make pictures or videos when they are working every day. The inertial sensor-based gait data collection is a recent addition to the gait analysis field. We have identified the age of people in this paper from an inertial sensor-data. We obtained the gait data from the University of Osaka. Convolution Neural Network (CNN) and LSTM-Based Convolution Neural Network (LSTM-CNN) are two deep learning algorithms that have been used to predict people's ages. The accuracy of age prediction via CNN is around 71.45%, while it is around 65.53% via CNN-LSTM, according to the experimental results.