This document briefly describes the system submitted by the Speech Processing Lab of Instituto de Telecomunicações, pole of Coimbra (SPL-IT) to the Query by Example Search on Speech Task (QUESST) of MediaEval 2014. Our approach is based on merging results of a phoneme recognition system using three different languages. A version of Dynamic Time Warping (DTW) using posteriorgram distances was cr...

This paper presents a robust voicing estimation algorithm for low bit rate harmonic speech coding. The algorithm is based on waveform time-warping followed by spectral matching based on voiced and unvoiced local spectral models. The objective of time warping is to reduce the effect of pitch variations on the voicing decision. Several adaptive techniques are used to improve the flexibility and r...

Retrieving the lyrics of a sung recording from a database of text documents is a research topic that has not received attention so far. Such a retrieval system has many practical applications, e.g. for karaoke applications or for indexing large song databases by their lyric content. In this paper, we present such a lyrics retrieval system. In a first step, phoneme posteriorgrams are extracted f...

This paper presents the results of applying gait and activity recognition on a commercially available mobile smartphone, where both data collection and real-time analysis was done on the phone. The collected data was also transferred to a computer for further analysis and comparison of various distance metrics and machine learning techniques. In our experiment 5 users created each 3 templates o...

In the paper, a new method for computing weighted distances for Dynamic Time Warping based speaker verification systems is described. Weighted distances use coefficients determined usually globally and this, of course, does not consider the phonetic content of the vocal pattern. The goal of local weighting is to connect the computation of the weights to the phonetic events occurring in the patt...

While voice conversion methods have been popularly applied to convert the speech signals uttered by a source speaker to a target speaker, frame-based voice conversion generally suffers from incorrect alignment using only spectral distance and therefore generate improper conversion results. In a parallel phone sequence, the alignment using minimum spectral distance between frame-based feature ve...

This study proposes a two-pass spoken term detection (STD) method. The first pass uses a phoneme-based dynamic time warping (DTW)-based STD, and the second pass recomputes detection scores produced by the first pass using conditional random fields (CRF)-based triphone detectors. In the second-pass, we treat STD as a sequence labeling problem. We use CRF-based triphone detection models based on ...

This study presents a new online signature verification system based on fuzzy modelling of shape and dynamic features extracted from online signature data. Instead of extracting these features from a signature, it is segmented at the points of geometric extrema followed by the feature extraction and fuzzy modelling of each segment thus obtained. A minimum distance alignment between the two samp...

In text-dependent speaker verification the speech signals have to be time-aligned. For that purpose dynamic time warping (DTW) can be used which performs the alignment by minimizing the Euclidean cepstral distance between the test and the reference utterance. While the cumulative Euclidean cepstral distance, which can be gathered from the DTW algorithm, could be used directly to discriminate be...

In this paper, we present the systems that the Natural Language Engineering and Pattern Recognition group (ELiRF) has submitted to the MediaEval 2013 Spoken Web Search task. All of them are based on a Subsequence Dynamic Time Warping algorithm and are zero-resources systems.