Jakub Gałka

Playback attack detection for text-dependent speaker verification over telephone channels☆

Abstract Playback attacks constitute one of the biggest threats in biometric speaker verification systems, in which a previously recorded passphrase is played back by an unprivileged person in order to gain access. This paper features a description of the playback attack detection (PAD) algorithm, designed to protect text-dependent speaker verification systems from the aforementioned spoofing attacks. The paper also describes the usage of spectral landmarks and score normalization methods in the playback detection algorithm. Different factors are discussed in terms of the performance of the algorithm. The authors investigate two issues: (1) extracting the PAD features which are robust against channel variations and (2) the robustness of the algorithm in adverse acoustical environments (e.g. office, street, cocktail party noise). The experiments are performed on a prepared speech corpus containing 4187 occurrences of a passphrase spoken by 175 speakers. The results of the experiment show the equal error rate (EER) to be as low as 1.0%. These findings demonstrate that such spoofing-oriented playback attacks can be effectively detected and should not be considered a significant argument against applications of text-dependent speaker verification.

Triphone Statistics for Polish Language

The Polish text corpus was analysed to find information about phoneme statistics. We were especially interested in triphones as they are commonly used in many speech processing applications like HTK speech recogniser. An attempt to create the full list of triphones for Polish language is presented. A vast amount of phonetically transcribed text was analysed to obtain the frequency of triphone occurrences. A distibution of frequency of triphones occuring and other phenomena are presented. The standard phonetic alphabet for Polish and methods of providing phonetic transcriptions are described.

Inertial Motion Sensing Glove for Sign Language Gesture Acquisition and Recognition

The most popular systems for automatic sign language recognition are based on vision. They are user-friendly, but very sensitive to changes in regard to recording conditions. This paper presents a description of the construction of a more robust system-an accelerometer glove-as well as its application in the recognition of sign language gestures. The basic data regarding inertial motion sensors and the design of the gesture acquisition system as well as project proposals are presented. The evaluation of the solution presents the results of the gesture recognition attempt by using a selected set of sign language gestures with a described method based on Hidden Markov Model (HMM) and parallel HMM approaches. The proposed usage of parallel HMM for sensor-fusion modeling reduced the equal error rate by more than 60%, while preserving 99.75% recognition accuracy.

Application of HTK to the Polish language

A speech recognition system based on HTK for Polish is presented. It was trained on 365 utterances, all spoken by 26 males. The features of Polish with respect to speech recognition are described. Some aspects of speech recognition differ in comparison to English. Errors in recognition were analysed in details in an attempt to find reasons and scenarios of wrong recognitions.

Voice authentication embedded solution for secured access control

The paper presents the concept of the embedded solution for voice biometric access system. The most important requirements for access control systems are presented, as well as the resulting design intent. The architecture of the system, its functionality and the methods used to verify the speakers is described along with a discussion of performance optimization methods of the implementation. Issues of voice biometrics and automatic speech processing are discussed as well.

WFT – Context-Sensitive Speech Signal Representation

Progress of automatic speech recognition systems’ (ASR) development is, inter alia, made by using signal representation sensitive for more and more sophisticated features. This paper is an overview of our investigation of the new context-sensitive speech signal’s representation, based on wavelet-Fourier transform (WFT), and proposal of it’s quality measures. The paper is divided into 5 sections, introducing as follows: phonetic-acoustic contextuality in speech, basics of WFT, WFT speech signal feature space, feature space quality measures and finally conclusion of our achievements.

Polish phones statistics

The phonemic statistics were collected from several large Polish corpora. The paper presents methodology of the acquisition process, summarisation of the data and some phenomena in the statistics. Triphone statistics apply context-dependent speech units which have an important role in speech technologies. The phonemic alphabet for Polish, SAMPA, and methods of providing phonemic transcriptions are described with detailed comments.

POLISH PHONEME STATISTICS OBTAINED ON LARGE SET OF WRITTEN TEXTS

The phonetical statistics were collected from several Polish corpora. The paper is a summaryof the data which are phoneme n-grams and some phenomena in the statistics. Triphonestatistics apply context-dependent speech units which have an important role in speech recognitionsystems and were never calculated for a large set of Polish written texts. The standardphonetic alphabet for Polish, SAMPA, and methods of providing phonetic transcriptions are described.

Wavelet Transform in Speech Segmentation

A non-uniform speech segmentation method based on discrete wavelet transform is used for the localization of phoneme boundaries. A vector of real values representing the digital speech signal is decomposed into phone-like units by placing segment borders according to the result of the multiresolution analysis. The final decision on localization of boundaries is taken by analysis of the energy flow among the decomposition levels. Distribution-like event functions indicate events, regarded as the segment boundaries.

Mean Best Basis Algorithm for Wavelet Speech Parameterization

In this paper, we propose a feature selection and transformation approach for universal steganalysis based on Genetic Algorithm (GA) and higher order statistics. We choose three types of typical statistics as candidate features and twelve kinds of basic functions as candidate transformations. The GA is utilized to select a subset of candidate features, a subset of candidate transformations and coefficients of the Logistic Regression Model for blind image steganalysis. The Logistic Regression Model is then used as the classifier. Experimental results show that the GA based approach increases the blind detection accuracy and also provides a good generality by identifying an untrained stego-algorithm. *