Research On Algorithm Based On Footstep Identification

Biological recognition technology combine high-tech means together, such as Optics, acoustics, biological sensors and biological statistical principles and so on. It identify personal identity by the inherent physical properties of humans(such as fingerprint, face, iris, etc.) and behavior characteristics, such as handwriting, voice, gait, etc.) Biological characteristics of humans are usually unique, measurable and inherited, which determines its great potential for development compared with traditional authentication technology.Footsteps recognition is a mew biometric identification technology, which takes advantage of the sound or vibration waveform features while a person is walking and can be used to estimate whether there is a footsteps. Footsteps based on the characteristics of acoustic and can be used to personal identity recognition. Normally individual footstep is different, with walking characteristics and the personal identity information contained, reliable and unique, which makes the footsteps identification of great development and application. Footsteps recognition has low requirements for acquisition device, convenient and effective collection as well as and high acceptance compared to other biometric technologies such as fingerprint identification, iris recognition and palmprint recognition and so on, which makes footstep have also got increasing concerns and become a new research hotspot in identity recognition field, especially of great significance in home surveillance, anti-theft security, military reconnaissance and other fields under a certain environment. Footstep can be also used to enable surveillance video, saving the video storage medium.In this paper, footsteps identification is described in details. Some innovations are proposed based on the key technologies of feature extraction and recognition as follows:1. In the preprocessing of footsteps, double threshold detection method is used for endpoint detection of footsteps; wiener filtering is used for footsteps with low noise in real environment, and compared with footsteps in quiet environment. A continuous footsteps, it is unable to locate noise as well as start of footsteps, which will affect the system recognition rate and calculation speed. So the endpoint detection will not only help to improve the recognition rate, but also greatly bring up the calculation speed. So far, footsteps are studied in quiet environment, but usually footstep identification is conducted in a noisy environment. Noise has a great influence on features based on acoustic parameters。 So de-noising is a necessary process in feature extraction and recognition and must be considered. In this paper, The method of wiener filtering is used for footsteps with noise in preprocessing.2. Footsteps of judgment, the likelihood of degrees in GMM(Gaussian Mixture Models) is put forward as a way to judge footsteps. in this method, Mel cepstrum coefficient (MFCC) is firstly used as a feature of footstep signal, and then expectation maximum (EM) algorithm is used to estimate the maximum value of logarithm likelihood function. logarithmic likelihood value can be drew through test, and its similarity degree of simulation can be measured by logarithmic likelihood value range. Results show that the recognition rate of this method was96%, commendable to judge footsteps.3. Feature extraction and recognition of Footsteps, the existing research methods mostly use acoustic parameters as features, such as MEL cepstrum coefficient, envelope spectrum similarity and so on, which is sensitive to mechanisms of different voices. has great constraints and restrictions for the same person in different shoes or walking on different floors based on which we put forward a new method for feature extraction, using footstep duration and interval time as features and then recognized by k-Nearest Neighbor algorithm (K-NN). The experimental results show that our method is effective for footsteps identification, more robust and adaptable for footsteps under different acoustic mechanism.