Wavelet-based cepstrum calculation

In this paper we present a new wavelet-based algorithm for low-cost computation of the cepstrum. It can be used for real time precise pitch determination in automatic speech and speaker recognition systems. Many wavelet families are examined to determine the one that works best. The results confirm the efficacy and accuracy of the proposed technique for pitch extraction. (C) 2008 Elsevier B.V. All rights reserved.

Nosturivaihteiden kunnonvalvontajärjestelmän kehittäminen värähtelymittauksia käyttäen

Työssä selvitettiin värähtelymittauksien käyttömahdollisuuksia nosturivaihteiden ennakoivassa kunnonvalvonnassa. Työ on painottunut lähinnä värähtelymittausten mittaus- ja analyysiteknisten näkökohtien tarkasteluun ja toimii niiltä osin al-kuselvityksenä värähtelymittauspohjaisen kunnonvalvontajärjestelmän luomiselle. Kirjallisuuden mukaan yleisimpiä ja käyttökelpoisimpia värähtelyanalyysejä vaihteistojen kunnonvalvonnassa ovat spektri-, kepstri-, ja demodulointitekniik-kaan perustuvat analyysit, joita kokeiltiin vauriokokeilla. Työssä käsitellään myös iskusysäysmenetelmää ja aikatason tunnuslukuja. Vauriokokeilla saatiin tietoa hammasvaurioiden havaitsemisesta värähtelymittauksilla, mutta laakerivaurioista kaivataan sen sijaan vielä lisää tietoa. Hammasvaurioiden tarkastelussa spektri-analyysin ohella PeakVue -analyysi osoittautui tehokkaaksi, kun siinä käytettiin sopivaa suodatusaluetta. Lisäksi työhön on koottu kokemuksia nosturivaihteiden käytännön värähtelymittauksista ja annettu ohjeita niiden suorittamisesta. Työn tulokset osoittavat, että värähtelymittauspohjaisella kunnonvalvontajärjes-telmällä voidaan parantaa nostovaihteiden ennakoivaa kunnonvalvontaa. Nosto-vaihteiden värähtelymittauksia kannattaa siis ehdottomasti jatkaa ja kunnonval-vontajärjestelmää edelleen kehittää. Siirtovaihteiden valvonta värähtelymittauksilla on sen sijaan hankalaa niiden epävakaan kuormitustilan vuoksi.

Speech Recognition of Malayalam Numbers

Digit speech recognition is important in many applications such as automatic data entry, PIN entry, voice dialing telephone, automated banking system, etc. This paper presents speaker independent speech recognition system for Malayalam digits. The system employs Mel frequency cepstrum coefficient (MFCC) as feature for signal processing and Hidden Markov model (HMM) for recognition. The system is trained with 21 male and female voices in the age group of 20 to 40 years and there was 98.5% word recognition accuracy (94.8% sentence recognition accuracy) on a test set of continuous digit recognition task.

Acoustic characterisation of cinema screens

As a consequence of cinema screens being placed in front of screen-speakers, a reduction in sound quality has been noticed. Cinema screens not only let the sound go through them, but also absorb a small amount of it and reflect the sound which impacts on the screen to the back, coming forward again in case it impacts on the loudspeaker. This backwards reflection in addition to the signal coming from the loudspeaker can lead to constructive or destructive interference at certain frequencies which usually results in comb filtering. In this project, this effect has been studied through researching amongst various data sheet provided by different manufacturers, acoustical measurements completed in the large anechoic chamber of the ISVR and some theoretical models developed with MatLab software. If results obtained with MatLab are accurate enough in comparison to the real measurements taken in the anechoic chamber this would lead to a good way to predict which would be the attenuation added to the system at each frequency, given that not all manufacturers provide an attenuation curve, but only an average attenuation. This average attenuation might be useless as sound waves have different wavelengths and its propagation through partitions varies. In fact, sound is composed by high and low frequencies, where high frequencies are characterised by a small wavelength which is usually easier to attenuate than low frequencies that characterised by bigger wavelengths. Furthermore, this information would be of great value to both screen manufacturers, who could offer a much more precise data in their data sheets; and customers, who would have a great amount of information to their disposal before purchasing and installing anything in their cinemas, being able to know by themselves which screen or loudspeaker should be best to meet their expectative. RESUMEN. La aparición de la digitalización de las bandas sonoras para las películas hace posible la mejora en la calidad de sonido de los cines. Sin embargo, un aspecto a tener en cuenta en esta calidad del sonido es la transmisión de éste a través de la pantalla, ya que normalmente tras ella se encuentran situados los altavoces. Las propiedades acústicas varían dependiendo del tipo de pantalla que se utilice, además de haber poca información a la que acceder para poder valorar su comportamiento. A lo largo de este proyecto, se analizan tres muestras de pantallas distintas donadas por distintos fabricantes para poder llegar a la conclusión de dependiendo del tipo de pantalla cuál es la distancia óptima a la que localizar la pantalla respecto al altavoz y con qué inclinación. Dicho análisis se realizó en la cámara anecoica del ISVR (University of Southampton) mediante la construcción de un marco de madera de 2x2 m en el que tensar las pantallas de cine, y un altavoz cuyo comportamiento sea el más similar al de los altavoces de pantalla reales. Los datos se captaron mediante cuatro micrófonos colocados en posiciones distintas y conectados al software Pulse de Brüel & Kjær, a través del cual se obtuvieron las respuestas en frecuencia del altavoz sin pantalla y con ella a diferentes distancias del altavoz. Posteriormente, los datos se analizaron con MatLab donde se calculó la atenuación, el factor de transmisión de la presión (PTF) y el análisis cepstrum. Finalmente, se realizó un modelo teórico del comportamiento de las pantallas perforadas basado en las placas perforadas utilizadas para atenuar el sonido entre distintas habitaciones. Como conclusión se llegó a que las pantallas curvadas son acústicamente más transparentes que las pantallas perforadas que a partir de 6 kHz son más acústicamente opacas. En las pantallas perforadas la atenuación depende del número de perforaciones por unidad de área y el diámetro de éstas. Dicha atenuación se reducirá si se reduce el diámetro de las perforaciones de la pantalla, o si se incrementa la cantidad de perforaciones. Acerca del efecto filtro peine, para obtener la mínima amplitud de éste la pantalla se deberá situar a una distancia entre 15 y 30 cm del altavoz, encontrando a la distancia de 30 cm que la última reflexión analizada a través de Cepstrum llega 5 ms más tarde que la señal directa, por lo cual no debería dañar el sonido ni la claridad del habla.

Aplicação de técnicas de classificação em séries temporais

Hoje, há um interesse crescente pela aprendizagem de uma segunda língua, quer seja por razões profissionais ou pessoais. Esta é uma tendência que se vai afirmando num mundo cada vez mais interconectado. Por outro lado, a democratização das tecnologias computacionais torna possível pensar em desenvolver novas técnicas de ensino de línguas mais automatizadas e personalizadas. Esta dissertação teve como objetivo estudar e implementar um conjunto de técnicas de processamento de sinal e de classificação de séries temporais úteis para o desenvolvimento de metodologias do ensino oral com feedback automático. São apresentados resultados preliminares sobre a prestação destas técnicas, e avaliada a viabilidade deste tipo de abordagem.

Electroencephalogram cepstral distances in alzheimer’s disease diagnosis

Alzheimer's disease (AD) represents one ofthe greatest public health challenges worldwide nowadays, because it affects millions of people ali o ver the world and it is expected that the disease will increase considerably in the near future. This study is the first application attempt of cepstral analysis on Electroencephalogram (EEG) signals to find new parameters in arder to achieve a better differentiation belween EEGs of AD patients and Control subjects. The results show that the methodology that uses a combined Wavelet (WT) Biorthogonal (Bior) 3.5 and cepstrum analysis was able to describe the EEG dynamics with a higher discriminative power than the other WTs/spectmm methodologies m previous studies. The most important significance figures were found in cepstral distances between cepstrums oftheta and alpha bands (p=0. 00006<0. 05).