Speech Signal In Time Domain Research Articles

An efficient speech perceptual hashing authentication algorithm based on DWT and symmetric ternary string

According to the situation that speech perceptual hashing methods are not appropriated for real-time speech content authentication in mobile computing environment, a novel DWT-based perceptual hashing algorithm, which uses a combination of time-domain and frequency-domain features, was proposed to protect the speech data in the cloud. Firstly, by discrete wavelet transform (DWT), a new signal in frequency-domain is generated from the original speech signal after pre-processing and intensity-loudness transform (ILT). Secondly, coefficients of low frequency wavelet decomposition are partitioned into equal-sized and non-overlapping blocks, and logarithmic short-time energy of each block is computed to obtain speech signals features in frequency-domain. Finally, combined with spectral flux features (SFF) of speech signal in time-domain, a ternary perceptual hashing sequence is created. Experiment results illustrate that ternary form is better to stand for hash digest than binary form, the proposed algorithm has a good robustness against content preserving operations, discrimination, good compaction and high efficiency, and detects the tamper localisation as well.

Frame Selection for Robust Speaker Identification: A Hybrid Approach

Identification of a person using voice is a challenging task under environmental noises. Important and reliable frame selection for feature extraction from the time-domain speech signal under noise can play a significant role in improving speaker identification accuracy. Therefore, this paper presents a frame selection method using hybrid technique, which combines two techniques, namely, voice activity detection (VAD) and variable frame rate (VFR) analysis. It efficiently captures the active speech part, the changes in the temporal characteristics of the speech signal, taking into account the signal-to-noise ratio, and thereby speaker-specific information. Experimental results on noisy speech, generated by artificially adding various noise signals to the clean YOHO speech at different SNRs have shown improved results for the frame selection by the hybrid technique in comparison with any one of the techniques used for the hybrid. The proposed hybrid technique outperformed both the VFR and the widely used Gaussian statistical model based VAD method for all noise scenarios at different SNRs, except for the Babble noise corrupted speech at 5 dB SNR, for which, VFR performed better. Considering the average identification accuracies of different noise scenarios, a relative improvement of 9.79% over the VFR, and 18.05% over the Gaussian statistical model based VAD method has been achieved.

Estimating acoustic speech features in low signal-to-noise ratios using a statistical framework

Accurate estimation of acoustic speech features from noisy speech and from different speakers is an ongoing problem in speech processing. Many methods have been proposed to estimate acoustic features but errors increase as signal-to-noise ratios fall. This work proposes a robust statistical framework to estimate an acoustic speech vector (comprising voicing, fundamental frequency and spectral envelope) from an intermediate feature that is extracted from a noisy time-domain speech signal. The initial approach is accurate in clean conditions but deteriorates in noise and with changing speaker. Adaptation methods are then developed to adjust the acoustic models to the noise conditions and speaker. Evaluations are carried out in stationary and nonstationary noises and at SNRs from −5 dB to clean conditions. Comparison with conventional methods of estimating fundamental frequency, voicing and spectral envelope reveals the proposed framework to have lowest errors in all conditions tested.

Wavelet Speech Enhancement Based on Nonnegative Matrix Factorization

For most of the state-of-the-art speech enhancement techniques, a spectrogram is usually preferred than the respective time-domain raw data since it reveals more compact presentation together with conspicuous temporal information over a long time span. However, the short-time Fourier transform (STFT) that creates the spectrogram in general distorts the original signal and thereby limits the capability of the associated speech enhancement techniques. In this study, we propose a novel speech enhancement method that adopts the algorithms of discrete wavelet packet transform (DWPT) and nonnegative matrix factorization (NMF) in order to conquer the aforementioned limitation. In brief, the DWPT is first applied to split a time-domain speech signal into a series of subband signals without introducing any distortion. Then we exploit NMF to highlight the speech component for each subband. Finally, the enhanced subband signals are joined together via the inverse DWPT to reconstruct a noise-reduced signal in time domain. We evaluate the proposed DWPT-NMF based speech enhancement method on the MHINT task. Experimental results show that this new method behaves very well in prompting speech quality and intelligibility and it outperforms the convnenitional STFT-NMF based method.

基于Labview的声卡数据采集与处理系统设计与实现

随着现代计算机科学技术的不断发展，labview图形编程开发技术已经被广泛应用。针对数据采集卡采集语音信号数据成本过高和计算机声卡性价比高等优点设计了基于labview的声卡数据采集处理系统。该系统可以根据用户需要设置合适的采样参数对采集到的语音信号进行时域和频域上的分析。该系统具有界面设计简单，扩展性好，功能多样化和数据处理能力强等优点。 With the development of modern computer science and technology, Labview graphical program-ming technology has been widely used nowadays. Considering the cost of the data acquisition card is high and the computer sound card is cost-effective, data acquisition card can be replaced by computer sound card. According to this we designed a sound card based on Labview data acquisi-tion processing system. This system can meet users’ different demands. We can do research on speech signal in time domain and frequency domain by setting appropriate sampling parameters. This system’s interface is simple and it has good expansibility, functional diversity and strong data processing capability.

Multi-Resolution Speech Spectrogram

An important aid in analysis & display of speech is sound spectrogram. It represents time-frequency-intensity display of short time spectrum. The quality of speech can be studied by visual inspection of spectrogram. This is one of the important applications of spectrogram in speech processing especially in speech enhancement. Another application of spectrogram is in isolating voiced and unvoiced regions. But to conclude from visual inspection the clarity of spectrogram is also important. Before plotting the spectrogram the time domain speech signal is converted to frequency domain. The transform domain used plays vital role in resolution of spectrogram. Generally Fast Fourier Transform is used to convert the time domain signal into frequency domain signal. This paper discusses the effect of using different transform for converting the time domain speech signal into frequency domain before plotting spectrogram. . It is observed that resolution of speech spectrogram is transform dependent.

Searching-and-averaging method of underdetermined blind speech signal separation in time domain

Underdetermined blind signal separation (BSS) (with fewer observed mixtures than sources) is discussed. A novel searching-and-averaging method in time domain (SAMTD) is proposed. It can solve a kind of problems that are very hard to solve by using sparse representation in frequency domain. Bypassing the disadvantages of traditional clustering (e.g., K-means or potential-function clustering), the durative- sparsity of a speech signal in time domain is used. To recover the mixing matrix, our method deletes those samples, which are not in the same or inverse direction of the basis vectors. To recover the sources, an improved geometric approach to overcomplete ICA (Independent Component Analysis) is presented. Several speech signal experiments demonstrate the good performance of the proposed method.

Clean speech reconstruction from MFCC vectors and fundamental frequency using an integrated front-end

The aim of this work is to enable a noise-free time-domain speech signal to be reconstructed from a stream of MFCC vectors and fundamental frequency and voicing estimates, such as may be received in a distributed speech recognition system. To facilitate reconstruction, both a sinusoidal model and a source-filter model of speech are compared by listening tests and spectrogram analysis, with the result that the former provides higher quality speech reconstruction. Analysis of the sinusoidal model shows that for clean speech reconstruction, both a noise-free spectral envelope and a robust estimate of the fundamental frequency and voicing are necessary. Investigation into fundamental frequency estimation reveals that an auditory model based approach gives superior performance over other methods of estimation. This leads to the proposal of an integrated front-end which uses the auditory model for both fundamental frequency and voicing estimation, and as the filterbank stage in MFCC extraction, and thereby reduces computation. Applying spectral subtraction to the auditory model parameters improves the spectral envelope estimates needed for clean speech reconstruction. Experiments on the Aurora connected digits database show that the auditory model-based MFCCs give comparable performance to that attained with conventional MFCCs. Speech reconstruction tests reveal that the combination of robust fundamental frequency and voicing estimation with spectral subtraction in the integrated front-end leads to intelligible and relatively noise-free speech.

Improved Data Modeling for Text-Dependent Speaker Recognition Using Sub-Band Processing

A growing body of recent work documents the potential benefits of sub-band processing over wideband processing in automatic speech recognition and, less usually, speaker recognition. It is often found that the sub-band approach delivers performance improvements (especially in the presence of noise), but not always so. This raises the question of precisely when and how sub-band processing might be advantageous, which is difficult to answer because there is as yet only a rudimentary theoretical framework guiding this work. We describe a simple sub-band speaker recognition system designed to facilitate experimentation aimed at increasing understanding of the approach. This splits the time-domain speech signal into 16 sub-bands using a bank of second-order filters spaced on the psychophysical mel scale. Each sub-band has its own separate cepstral-based recognition system, the outputs of which are combined using the sum rule to produce a final decision. We find that sub-band processing leads to worthwhile reductions in both the verification and identification error rates relative to the wideband system, decreasing the identification error rate from 3.33% to 0.56% and equal error rate for verification by approximately 50% for clean speech. The hypothesis is advanced that, unlike the wideband system, sub-band processing effectively constrains the free parameters of the speaker models to be more uniformly deployed across frequency: as such, it offers a practical solution to the bias/variance dilemma of data modeling. Much remains to be done to explore fully the new paradigm of sub-band processing. Accordingly, several avenues for future work are identified. In particular, we aim to explore the hypothesis of a practical solution to the bias/variance dilemma in more depth.

Estimation of articulatory movement and its application to speech synthesis

In this paper, two methods are investigated to estimate the movement of the vocal tract modeled by a three-dimensional articulator described by seven physical parameters. First, we use the cubic spline method to interpolate the articulatory parameters between consecutive phonemes. Second, the articulatory parameters are piecewise linearly interpolated and then passed by a cubic spline smoothing filter. The cubic spline smoothing filter is a low-pass filter with the maximum flatness property. The filter bandwidth can be easily adjusted by a control parameter. A graphical animation is built to visualize the articulatory movement of phonetic strings and to compare the results from the two different schemes. Some English phonemes are synthesized based on the articulatory model and the governing Webster equation in frequency domain considering viscous losses. An overlap-and-add method is used to convert the frequency domain spectrum to time domain speech signal. Finally, some example of the synthesized speech phonemes will be demonstrated. [We want to thank the National Science Foundation (NSF) for the support of our work.]

Comparison of several speech signal feature parameters for automatic speech recognition

This paper is devoted to the problem of the selection of those speech signal features which are relevant for automatic speech recognition. Discrete power spectrum features, i.e. the sign and rank-order functions of a bandpass filter output are analyzed together with more standard features such as LPC coefficients and the short-time spectrum measured by means of a bandpass filter bank. Structural time-domain speech signal features are also analyzed. A comparison of the features was performed using a non parametric (template matching) model of an isolated word recognition system. The experiments were based on Serbo-Croat digits spoken by 109 speakers. The recognition accuracy of each of the speech signal features was established. The structure of recognition errors and their causes was also studied in relation to the test vocabulary.