A double-talk detection algorithm using a psychoacoustic auditory model

An acoustic echo cancellation system is one of the most important breakthrough in the field of adaptive systems. Today acoustic echo cancellers (AEC) are an integral part of full duplex hands-free voice communication. Conventional echo cancellers use a linear model to represent the echo path. However many consumer devices include loud-speakers and power amplifiers that generate non-linear distortions.Non-linearity occurs due to the use of low cost electronic loud speakers, microphones and poorly designed enclosures in an AEC system. Non-linearity causes vibration and harmonic distortion and also degrades the speech quality. Double talk detector (DTD) is a key component of an AEC. A DTD is used to sense when the far end signal is corrupted by the near end speech. The DTD freezes the adaptation of model filter to prevent the divergence of the adaptive filter. Various authors have proposed different algorithms for double talk detection. Some of the most popular algorithms are Geigel algorithm, cross-correlation based DTD, normalized cross correlation based DTD, variable impulse response DTD etc. In this paper several double talk detection algorithms in a non-linear platform of an AEC has been discussed.

Echo cancellers typically employ a doubletalk detection (DTD) algorithm in order to keep the adaptive filter from diverging in the presence of near-end speech signal or other disruptive sounds in the microphone signal. A novel doubletalk detection algorithm based on techniques similar to those used for audio signal watermarking was introduced by the authors. The application of the described DTD algorithm within acoustic echo cancellation system is presented. The comparison of the proposed algorithm with very common, but simple Geigel algorithm and representing current state-of-the-art Normalized Cross-Correlation algorithms is performed. Both objective (ROC) and subjective (listening tests) performance evaluation methods are employed to obtain exhaustive evaluation results in simulated real-world conditions. The evaluation results are presented and their relevance is discussed. An issue of algorithms' computational complexity is emphasized and conclusions are drawn.

this paper presents a novel acoustic echo canceller (AEC) consisting of an adaptive FIR filter and a double-talk detector (DTD). By employing the widely used LPC-based speech coder for decorrelation, we develop a variable step-size decorrelation NLMS algorithm for the adaptive filter and an echo-path mismatch detection algorithm for the double-talk detector. This new approach reduces the computational complexity of both adaptive algorithm and double-talk detection algorithm in order of magnitude. Simulation results show that the proposed algorithm has a superior performance compared with other existing relative algorithms.

Acoustic Echo Cancellation (AEC) is employed in hands-free communication systems to enhance communication experience. Double-Talk Detection (DTD) plays a vital role in AEC to distinguish the far-end and near-end signals and its failure will cause the system divergence. An efficient new DTD algorithm based on the joint signal energy and cross-correlation estimation techniques is proposed in this paper to reduce the DTD faults. The proposed algorithm can provide fast DTD and enhanced system performance without increasing much more computational complexity over the conventional techniques. Simulation results show the performance improvement achieved by the proposed technique.

Echo cancelers commonly employ a doubletalk detector (DTD), which is essential to keep the adaptive filter from diverging in the presence of near-end speech and other disruptive noise. There have been numerous algorithms to detect doubletalk in an acoustic echo canceler (AEC). In those applications, typically, the threshold is chosen only by some heuristic method and the performance evaluation is very subjective. In this study, we develop a way to objectively evaluate DTD algorithms based on the standard statistical methods of detection theory. A receiver operating characteristic (ROC) is derived to characterize DTD performance. Several DTD algorithms are examined and simulated under typical real-world operating conditions using measured room responses and signals taken from a digital speech database. The DTD methods are then evaluated and compared using the ROC metric.

When an adaptive filter is used for echo cancellation, it is essential to prevent the filter from diverging in situations when the echo signal is contaminated with near-end disturbance, i.e., during double-talk. This paper presents an extension of a previously proposed double-talk detector for improved performance. It is shown that the computational complexity of the proposed detector is lower than that of the well-used normalized cross correlation (NCC) double-talk detector, at the cost of performance. Further, it is shown that there can be a significant performance difference, in terms of detecting double-talk, between having a fixed echo cancellation filter, which is a common strategy in objective evaluation techniques, and an adaptive filter, which is more close to realistic conditions.

In this paper, a single channel Kalman filtering in the spectral domain is proposed for acoustic echo and noise cancellation (AENC). An augmented extended Kalman filter is used in this work which is suitable for processing general complex-valued nonlinear and nonstationary signals. The near-end speech and background noise are modelled to make the filter robust against double-talk and against noise. In this work, the approach for echo cancellation does not depend on double-talk detector (DTD) to model the echo path, when signals other than the echoed far-end signal are present. The performance evaluations are conducted over the NOIZEUS corpus. The experimental results demonstrate that the ideal Kalman filter (with near-end speech parameters) outperforms other enhancement methods of echo cancellation in terms of subjective and objective measures. The non-ideal Kalman filter (when near-end speech parameters are obtained by an EM algorithm) also shows promising results after the ideal Kalman filter.

In this paper, the adaptive cancellation structure is first developed based on the LMS algorithm and FIR adaptive filtering. Then the novel practical noise and echo cancellation systems are built using the proposed adaptive technique and implemented using TX320TMS67C13 DSKs, which are Texas Instruments’ Digital Signal Processing (TI DSP) boards. Although adaptive filtering is an exciting topic in which many real-life applications can be explored [1]-[6], [9], building such a real-time system is often challenging due to the use of theoretical math, advanced DSP knowledge and practical industrial hands-on experience [1],[4]-[6],[9]. Therefore, this paper indicates that it is possible to apply traditional mathematics in adaptive filtering theory to real-time practical DSP systems. With the MATLAB software tool, we can simulate and verify various adaptive filtering designs first. Then, development and implementation of different noise or echo cancellation systems with adaptive filtering techniques can be successfully performed using the floating-point digital signal processor, TX320TMS67C13 DSK. Furthermore, it can be shown that TX320TMS67C13 DSKs with their stereo channels offer more effective and flexible tools for various noise cancellation applications.

Acoustic Echo Cancelation is a typical event in today's telecommunication fields. The input speech signal interference caused by acoustic echo is disturbing for both users side and causes a reduction in the quality of the original signal. The adaptive filtering techniques are used to reduce the unwanted echo, and increasing quality of the speech signal. This Paper focus on different Adaptive filters algorithms and characterized by the Normalized Least Mean Square (NLMS) and Variable Step Size Least Mean Square (VSSLMS) and also This Paper actualizes Real Time Acoustic Echo Cancellation on Tms320c6713dsk. The primary modules of this reverberation canceller are a versatile channel framework utilizing Normalized Least Mean Square (NLMS) calculation, a Double-Talk Detector (DTD) and a Non-Linear Processor (NLP). The Setup of NLMS based versatile framework, Modified NLMS with Variable Step Size calculations and Frequency Domain based Fast LMS algorithms are to be actualized on the DSP equipment. This implementation shows fast response, speed, convergence of time and good tracking capability. The calculation of different adaptive algorithms must be analysed utilizing MATLAB.

To attain good performance in an acoustic echo cancellation system, it is important to have a variable step size (VSS) algorithm as part of an adaptive filter. In this paper, we are concerned with the development of a VSS algorithm for a recently proposed subband affine projection (SAP) adaptive filter. Two popular VSS algorithms in the literature are the methods of delayed coefficients (DC) and variable regularization (VR). However, the merits and demerits of them are mutually exclusive. We propose a VSS algorithm* that is a hybrid of both methods and combines their advantages. An extensive study of the new algorithm in different scenarios like the presence double-talk (DT) during the transient phase of the adaptive filter, DT during steady state, and varying DT power is conducted and reasoning is given to support the observed behavior. The importance of the method of VR as part of a VSS algorithm is emphasized.

Most of the echo cancellers are equipped with a double-talk detector (DTD) in order to control the behavior of the adaptive filter during double-talk situations. In this paper, we propose a class of DTDs based on the Holder inequality. These DTDs are simple to implement, have low computational complexity, and perform well even for low echo-to-noise ratios. As a particular case, it is shown that the well-known Geigel algorithm can be obtained from this approach.

Doubletalk detection is an important part of a practical echo canceller implementation, but a difficult problem is calibrating the doubletalk detector for arbitrary environments and input signals. In this paper, it is shown that a statistical model of a doubletalk detection variable's probability density function (PDF) can be used to obtain an optimal detection threshold and expected detection performance curves. In particular, a statistical analysis of a recently proposed cross-correlation-based doubletalk detector is presented. The doubletalk detection variable is modeled in terms of its constituent parameter estimators, resulting in conditional PDFs in the absence and presence of doubletalk. These are used to obtain a signal-adaptive detection threshold for calibration, and to provide expected doubletalk detection probability. Simulations are presented comparing the theoretical and measured detection probability compared to a fixed detection threshold for speech input and doubletalk signals. The results indicate a close agreement with the proposed model for moderate-to-high levels of doubletalk

Most teleconferencing conversations are conducted in the presence of Acoustic echoes. Typically an adaptive filter is used to cancel the echo, with a control device called the doubletalk detector which controls the adaptation. We derive a novel test statistic for the doubletalk detection based on the cross-correlation between the microphone signal and the cancellation error for the frequency domain adaptive algorithm. The main advantage of the proposed algorithm is its simplicity and computational efficiency. We compare our results with the normalized cross-correlation based doubletalk detector proposed. We also generalize the idea of the proposed doubletalk detector (single-channel) to the matrix case (multi-channel).

This paper presents a normalized cross-correlation- based doubletalk detector for use in a critically sampled subband acoustic echo canceller. The algorithm is derived from an existing doubletalk detector for fullband echo cancellers. The resulting per-subband doubletalk detectors are allowed to control adaptation independently in each subband. Computational complexity reduction techniques are described, along with a noise estimation scheme to compensate for the effects of background noise in hands-free and vehicular environments. Simulation results with narrowband speech doubletalk signals show that the proposed structure can achieve a higher level of convergence than a global doubletalk decision in the presence of doubletalk. Detection probability results are also presented.

Mobile phones used in cars require for safety reasons handset-free operation.This implies that a front-end is used which is able to reduce the ambient noise so that an acceptable intelligiblity is guaranteed for the far-end customer. Furthermore, the acoustic echo transmitted from the loudspeaker to the microphone has to be reduced so that it is not audible.Another aspect has to be taken into account: the system has to be robust, i.e. it has to be operable in a wide range of the signal-to-noise ratio down to 0 dB or less. Furthermore the demand for a-priori knowledge should be as low as possible because acquisition of knowledge induces time delays unacceptable for real-time systems and contradicts a low-cost solution which is required for a mass product.New developments in this paper concern the segmentation of the frequency band into subbands for all relevant components: noise reduction, echo cancellation and double-talk detection. Furthermore new concepts for echo reduction are presented.Key wordsRobustnessnoise reductionecho cancellation