Jiri Schimmel

Impact of vocal effort variability on automatic speech recognition

The impact of changes in a speakers vocal effort on the performance of automatic speech recognition has largely been overlooked by researchers and virtually no speech resources exist for the development and testing of speech recognizers at all vocal effort levels. This study deals with speech properties in the whole range of vocal modes - whispering, soft speech, normal speech, loud speech, and shouting. Fundamental acoustic and phonetic changes are documented. The impact of vocal effort variability on the performance of an isolated-word recognizer is shown and effective means of improving the systems robustness are tested. The proposed multiple model framework approach reaches a 50% relative reduction of word error rate compared to the baseline system. A new specialized speech database, BUT-VE1, is presented, which contains speech recordings of 13 speakers at 5 vocal effort levels with manual phonetic segmentation and sound pressure level calibration.

Real-Time Guitar Preamp Simulation Using Modified Blockwise Method and Approximations

The designing of algorithms for real-time digital simulation of analog effects and amplifiers brings two contradictory requirements: accuracy versus computational efficiency. In this paper, the simulation of a typical guitar tube preamp using an approximation of the solution of differential equations is discussed with regard to accuracy and computational complexity. The solution of circuit equations is precomputed and stored in N-D tables. The stored values are approximated, and therefore different approximation techniques are investigated as well. The approximated functions are used for output signal computation and also for circuit state update. The designed algorithm is compared to the numerical solution of the given preamp and also to the real preamp.

Surveillance and Quality Evaluation System for Real-Time Audio Signal Streaming

This paper deals with the design of an algorithm and application for an automatic system of real-time audio signal quality surveillance on personal computers. This system works without any knowledge of how the audio signal is processed. The audio signal quality and possible signal failures are monitored by being compared with the original signal in the time-frequency domain. The system supports several audio formats from mono to 5.1 and downmix for comparison between different audio formats. For high performance, SIMD technologies are used. The whole system consists of several local surveillance applications (clients) and a server application with SNMP client that monitors the local application states using the TCP/IP network.

Objective quality assessment for the acoustic zoom

Sound quality is a major issue of concern for all modern telecom and audiovisual systems. The evaluation of sound quality includes both objective and subjective methods. The subjective methods reflect the listeners perception of the sound and the objective methods try to evaluate the sound parameters affecting its quality in correlation with the results of the subjective methods. The subjective tests are time consuming, which is inconvenient during the development of new algorithms of audio signal processing and tuning its parameters. This article investigates the possibility of applying objective methods for audio quality assessment on an acoustic zooming system and it compares their results and the results of the subjective intelligibility test.

Objective evaluation of audible aliasing distortion in digital audio synthesis

This paper deals with aliasing distortion in digital audio signal synthesis of basic periodic waveforms with infinite Fourier series. There is aliasing when these signals are generated in digital domain because of its unlimited bandwidth. There are several synthesis techniques of these signals designed to avoid or reduce the aliasing distortion but they have high computing demands. Trivial generation of basic periodic waveforms is therefore still effective for sound synthesis when oversampling is used. There will always be the aliasing distortion but some spectral components produced by the aliasing will be masked with harmonic components and thus inaudible. This paper deals with objective evaluation of audible aliasing distortion with the help of a psychoacoustic model of simultaneous masking and compares the computing demands with other methods.

Performance Analysis of Parallel Signal Processing Algorithms in Audio Mixing Systems

Algorithms for real-time parallel processing of an audio signal in large-scale digital audio distribution networks, implemented on personal computer platform, are compared in this paper from the performance point of view. In such systems, the summing and multiplication of audio signal take up a significant portion of the processing power of the system. Therefore, there is a tendency to decrease its computing demands and thus reserve the computing power of the processor for other signal processing modules in the audio network. Various approaches can therefore be used to distribute computing among several threads. Seven approaches are analyzed in the paper and fastest one is found with regard to the size of audio sample buffers.

Spectral Characteristics of Nonlinear Systems Approximated by Piecewise Linear Function

This paper deals with an analytic solution of spectrum changes that appear in nonlinear discrete systems without memory whose transfer characteristics can be approximated via piecewise linear function. The paper also deals with relations between the mutual harmonics ratio and the approximation parameters. Furthermore, the dependence of the harmonics ratio on the amplitude of a harmonic input signal is presented for the most common systems whose transfer characteristics are approximated via piecewise linear function.

Surveillance System for Audio Broadcast

This paper deals with algorithm design for an automatic surveillance system for the stereophonic audio broadcast and cross-platform implementation of this system, using the virtual studio technology (VST) and the audio streaming input/output (ASIO) technologies for real-time digital audio signal processing on personal computers. First, the parameters of the stereophonic audio signal monitored are specified and then the results of analyses are presented, which are made by a variety of methods of digital signal processing. To monitor the audio broadcast parameters, the following time and frequency domain analyses were mainly used: RMS and peak analysis, correlation analysis, differential analyses, fast convolution, and delta-spectrum analysis. The implementation of the system using the VST and ASIO technologies is described in the end of the paper

Control of digital audio signal processing over communication networks

This paper deals with the design of a protocol for the control of digital audio signal processing over communication networks, primarily for TCP/IP networks, and its multi-platform parser. The application of this protocol in embedded audio processing units represents a modern approach to remote-controlled real-time processing of digital audio signals. The aim is to have the possibility to control audio signal processing in distributed processing networks over very long distances, e.g. via the Internet.