Improving visual noise insensitivity in small vocabulary audio visual speech recognition applications

Visual noise insensitivity is important to audio visual speech recognition (AVSR). Visual noise can take on a number of forms such as varying frame rate, occlusion, lighting or speaker variabilities. The use of a high dimensional secondary classifier on the word likelihood scores from both the audio and video modalities is investigated for the purposes of adaptive fusion. Preliminary results are presented demonstrating performance above the catastrophic fusion boundary for our confidence measure irrespective of the type of visual noise presented to it. Our experiments were restricted to small vocabulary applications.

Can audio-visual speech recognition outperform acoustically enhanced speech recognition in automotive environment?

The use of visual features in the form of lip movements to improve the performance of acoustic speech recognition has been shown to work well, particularly in noisy acoustic conditions. However, whether this technique can outperform speech recognition incorporating well-known acoustic enhancement techniques, such as spectral subtraction, or multi-channel beamforming is not known. This is an important question to be answered especially in an automotive environment, for the design of an efficient human-vehicle computer interface. We perform a variety of speech recognition experiments on a challenging automotive speech dataset and results show that synchronous HMM-based audio-visual fusion can outperform traditional single as well as multi-channel acoustic speech enhancement techniques. We also show that further improvement in recognition performance can be obtained by fusing speech-enhanced audio with the visual modality, demonstrating the complementary nature of the two robust speech recognition approaches.

Multiple cameras for audio-visual speech recognition in an automotive environment

Audio-visualspeechrecognition, or the combination of visual lip-reading with traditional acoustic speechrecognition, has been previously shown to provide a considerable improvement over acoustic-only approaches in noisy environments, such as that present in an automotive cabin. The research presented in this paper will extend upon the established audio-visualspeechrecognition literature to show that further improvements in speechrecognition accuracy can be obtained when multiple frontal or near-frontal views of a speaker's face are available. A series of visualspeechrecognition experiments using a four-stream visual synchronous hidden Markov model (SHMM) are conducted on the four-camera AVICAR automotiveaudio-visualspeech database. We study the relative contribution between the side and central orientated cameras in improving visualspeechrecognition accuracy. Finally combination of the four visual streams with a single audio stream in a five-stream SHMM demonstrates a relative improvement of over 56% in word recognition accuracy when compared to the acoustic-only approach in the noisiest conditions of the AVICAR database.

Automatic Audio Segmentation Using the Generalized Likelihood Ratio

This paper presents a novel technique for segmenting an audio stream into homogeneous regions according to speaker identities, background noise, music, environmental and channel conditions. Audio segmentation is useful in audio diarization systems, which aim to annotate an input audio stream with information that attributes temporal regions of the audio into their specific sources. The segmentation method introduced in this paper is performed using the Generalized Likelihood Ratio (GLR), computed between two adjacent sliding windows over preprocessed speech. This approach is inspired by the popular segmentation method proposed by the pioneering work of Chen and Gopalakrishnan, using the Bayesian Information Criterion (BIC) with an expanding search window. This paper will aim to identify and address the shortcomings associated with such an approach. The result obtained by the proposed segmentation strategy is evaluated on the 2002 Rich Transcription (RT-02) Evaluation dataset, and a miss rate of 19.47% and a false alarm rate of 16.94% is achieved at the optimal threshold.

Managing and analysing big audio data for environmental monitoring

Environmental monitoring is becoming critical as human activity and climate change place greater pressures on biodiversity, leading to an increasing need for data to make informed decisions. Acoustic sensors can help collect data across large areas for extended periods making them attractive in environmental monitoring. However, managing and analysing large volumes of environmental acoustic data is a great challenge and is consequently hindering the effective utilization of the big dataset collected. This paper presents an overview of our current techniques for collecting, storing and analysing large volumes of acoustic data efficiently, accurately, and cost-effectively.

Image processing and classification procedure for the analysis of Australian frog vocalisations

Frog species have been declining worldwide at unprecedented rates in the past decades. There are many reasons for this decline including pollution, habitat loss, and invasive species [1]. To preserve, protect, and restore frog biodiversity, it is important to monitor and assess frog species. In this paper, a novel method using image processing techniques for analyzing Australian frog vocalisations is proposed. An FFT is applied to audio data to produce a spectrogram. Then, acoustic events are detected and isolated into corresponding segments through image processing techniques applied to the spectrogram. For each segment, spectral peak tracks are extracted with selected seeds and a region growing technique is utilised to obtain the contour of each frog vocalisation. Based on spectral peak tracks and the contour of each frog vocalisation, six feature sets are extracted. Principal component analysis reduces each feature set down to six principal components which are tested for classification performance with a k-nearest neighbor classifier. This experiment tests the proposed method of classification on fourteen frog species which are geographically well distributed throughout Queensland, Australia. The experimental results show that the best average classification accuracy for the fourteen frog species can be up to 87%.

Application of image processing techniques for frog call classification

Frogs have received increasing attention due to their effectiveness for indicating the environment change. Therefore, it is important to monitor and assess frogs. With the development of sensor techniques, large volumes of audio data (including frog calls) have been collected and need to be analysed. After transforming the audio data into its spectrogram representation using short-time Fourier transform, the visual inspection of this representation motivates us to use image processing techniques for analysing audio data. Applying acoustic event detection (AED) method to spectrograms, acoustic events are firstly detected from which ridges are extracted. Three feature sets, Mel-frequency cepstral coefficients (MFCCs), AED feature set and ridge feature set, are then used for frog call classification with a support vector machine classifier. Fifteen frog species widely spread in Queensland, Australia, are selected to evaluate the proposed method. The experimental results show that ridge feature set can achieve an average classification accuracy of 74.73% which outperforms the MFCCs (38.99%) and AED feature set (67.78%).

A hybrid pre-whitening technique for detection of additive spread spectrum watermarks in audio signals

Pre-whitening techniques are employed in blind correlation detection of additive spread spectrum watermarks in audio signals to reduce the host signal interference. A direct deterministic whitening (DDW) scheme is derived in this paper from the frequency domain analysis of the time domain correlation process. Our experimental studies reveal that, the Savitzky-Golay Whitening (SGW), which is otherwise inferior to DDW technique, performs better when the audio signal is predominantly lowpass. The novelty of this paper lies in exploiting the complementary nature to the two whitening techniques to obtain a hybrid whitening (HbW) scheme. In the hybrid scheme the DDW and SGW techniques are selectively applied, based on short time spectral characteristics of the audio signal. The hybrid scheme extends the reliability of watermark detection to a wider range of audio signals.

Temporal Envelope Fit of Transient Audio Signals

We address the problem of temporal envelope modeling for transient audio signals. We propose the Gamma distribution function (GDF) as a suitable candidate for modeling the envelope keeping in view some of its interesting properties such as asymmetry, causality, near-optimal time-bandwidth product, controllability of rise and decay, etc. The problem of finding the parameters of the GDF becomes a nonlinear regression problem. We overcome the hurdle by using a logarithmic envelope fit, which reduces the problem to one of linear regression. The logarithmic transformation also has the feature of dynamic range compression. Since temporal envelopes of audio signals are not uniformly distributed, in order to compute the amplitude, we investigate the importance of various loss functions for regression. Based on synthesized data experiments, wherein we have a ground truth, and real-world signals, we observe that the least-squares technique gives reasonably accurate amplitude estimates compared with other loss functions.

Binaural Signal Processing Motivated Generalized Analytic Signal Construction and AM-FM Demodulation

Binaural hearing studies show that the auditory system uses the phase-difference information in the auditory stimuli for localization of a sound source. Motivated by this finding, we present a method for demodulation of amplitude-modulated-frequency-modulated (AM-FM) signals using a ignal and its arbitrary phase-shifted version. The demodulation is achieved using two allpass filters, whose impulse responses are related through the fractional Hilbert transform (FrHT). The allpass filters are obtained by cosine-modulation of a zero-phase flat-top prototype halfband lowpass filter. The outputs of the filters are combined to construct an analytic signal (AS) from which the AM and FM are estimated. We show that, under certain assumptions on the signal and the filter structures, the AM and FM can be obtained exactly. The AM-FM calculations are based on the quasi-eigenfunction approximation. We then extend the concept to the demodulation of multicomponent signals using uniform and non-uniform cosine-modulated filterbank (FB) structures consisting of flat bandpass filters, including the uniform cosine-modulated, equivalent rectangular bandwidth (ERB), and constant-Q filterbanks. We validate the theoretical calculations by considering application on synthesized AM-FM signals and compare the performance in presence of noise with three other multiband demodulation techniques, namely, the Teager-energy-based approach, the Gabor's AS approach, and the linear transduction filter approach. We also show demodulation results for real signals.

Robust noise reduction for speech and audio signals

Statistical model-based methods are presented for the reconstruction of autocorrelated signals in impulsive plus continuous noise environments. Signals are modelled as autoregressive and noise sources as discrete and continuous mixtures of Gaussians, allowing for robustness in highly impulsive and non-Gaussian environments. Markov Chain Monte Carlo methods are used for reconstruction of the corrupted waveforms within a Bayesian probabilistic framework and results are presented for contaminated voice and audio signals.

Towards a perceptually optimal spectral amplitude estimator for audio signal enhancement

We present a statistical model-based approach to signal enhancement in the case of additive broadband noise. Because broadband noise is localised in neither time nor frequency, its removal is one of the most pervasive and difficult signal enhancement tasks. In order to improve perceived signal quality, we take advantage of human perception and define a best estimate of the original signal in terms of a cost function incorporating perceptual optimality criteria. We derive the resultant signal estimator and implement it in a short-time spectral attenuation framework. Audio examples, references, and further information may be found at http://www-sigproc.eng.cam.ac.uk/~pjw47.