B.V. Uma

Low Complexity Speech Enhancement Algorithm for Improved Perception in Mobile Devices

In mobile phones, perceived quality of speech signal deteriorates significantly in the presence of background noise since near-end/surrounding noise also arrives at the near-end listener’s ears. The quality of the received signal varies widely depending upon signal strength and unavoidable background noise in the user environment. There is a need to improve the quality of received speech signal in noisy conditions by developing the speech enhancement algorithms. This paper focuses on the impact of the various background noises on signal perception and mechanisms to mitigate the noise impact for improved signal perception. Gain adjustment process with simple time domain approach has been adapted to improve the quality and intelligibility of the speech signal in the noisy environments by automatically increasing output levels when the noise dominates. Since time domain approach has been used, it is less complex and consumes very less battery power of the mobile and also efficiently masks the background noise. This paper studies the effect of various signal parameters on automatic gain adjustment of the received signal to enhance its perception.

Implementation of Novel Bit Loading Algorithm with Power Cut-Back in ADSL Transmitter on DSP

ADSL (Asymmetric Digital Subscriber Line) adopts modulation scheme called DMT (Discrete multi tone), a multi carrier system. The process of allocating bits into various sub-channels is called bit-loading. In this paper, a new bit-loading algorithm that reduces the power required to transmit data is proposed. Comparison on power cut back with other algorithms is presented. Also the implementation of the algorithm in the transmitter is discussed. A Matlab simulation model has been developed to simulate ADSL transmitter (with bit-loading), channel and receiver. The ADSL system with bit-loading algorithm is also implemented in hardware using TTs DSK6713 kit and comparison of software and hardware results are tabulated.

Performance enhancement and area optimization of 3×3 NoC using random arbiter

Network-on-Chip (NoC) is a new approach for designing the communication subsystem among IP cores in a System-on-Chip (SoC). NoC applies networking theory and related methods to on-chip communication and brings out notable improvements over conventional bus and crossbar interconnections. NoC offers a great improvement over the issues like scalability, productivity, power efficiency and signal integrity challenges of complex SoC design. In a NoC, the communication among different nodes is achieved by routing packets through a pre-designed network according to different routing algorithms. Therefore, architecture and related routing algorithm plays an important role to the improvement of overall performance of a NoC. The technique one which is used presently in node is priority based technique packet routing which leads the packet stacking which intern leads to performance degradation. In this paper, proposes a modified random Arbiter combined with deterministic XY routing algorithm to be used on router of NoC. In this method router contains random arbiter along with priority encoder which results into fast way to transfer packet via a specific path between the nodes of the network without stacking. This in turn optimizes the packet storage area and avoids collision because node arbiter will service the packets randomly without any priority. In addition to that this method will ensure a packet always reaches the destination through the possible shortest path without deadlock and livelock.

Speech enhancement to overcome the effect of near-end noise in mobile phones using psychoacoustics

In mobile phones, perceived quality of speech signal deteriorates significantly in the presence of near-end/surrounding noise as it arrives at the near-end listeners ears. There is a need to improve the intelligibility and quality of the received speech signal in noisy environments by developing speech enhancement algorithms. Paper focuses on the impact of near-end noises on signal degradation. Gain adjustment process employing psychoacoustic has been adopted to improve the intelligibility and quality of the speech signal in the noisy environments by automatically enhancing the speech signals when the noise dominates. Intelligibility and quality of enhanced speech signal are measured using SII and PESQ. Experimental results show intelligibility and quality improvement of speech signal with the proposed method over the unprocessed far-end speech signal. Approach is more efficient in overcoming the degradation of speech signals in presence of near-end noise.

Platform level design for Network on Chips

As technology improves, it is possible to integrate more number of transistors on a single die, which means it is possible to design any complex system on single chip. The networking in such system is very difficult as compared to less complex systems one which uses switching and bus technique. In order to overcome this, a special networking technique is used which is Network on Chip. In this paper, a technique is proposed to develop the platform level design to receive the error-free packets which improves the performance of Network on Chip. By receiving the error free packets, there is no need for the network to check the packet once again for correctness thereby reducing the time taken for the packet to reach from source to destination resulting in an increase in efficiency. This helps in improving the performance since the error packet will be eliminated at the platform level before sending it to the cluster level. The packets are stored into RAM and are classified with the help of packet classifier. The packet classifier will send the packets to respective cluster and node agents at lower hierarchical level. This will improve the performance by avoiding the erroneous packets in network.

Application of wavelet for improvement of rate-reach performance in ADSL interference environment

This paper discusses the cancellation of near end crosstalk (NEXT) and far end crosstalk (FEXT) using a non-parametric wavelet denoising technique. Crosstalk noise is estimated using heuristic Steinpsilas unbiased risk estimate (SURE) and Bayesian risk minimization threshold function. The noise level estimated by a threshold function is used by thresholding rules to mitigate the effect of noise in the signal. A comparison of rates achieved with three thresholding techniques viz; soft thresholds, hard thresholds and a garrote threshold have been presented. Results obtained with AWGN, 24 ADSL NEXT and 24 ADSL FEXT disturbers show a significant improvement in rate reach performance upwards of 4 Mbps. The computational complexity in the wavelet denoising technique is independent of the number of crosstalk sources and hence can be suitable candidate for a practical deployment of mitigation of cross talk.

Random arbiter and platform level design for improving the performance on 4×4 NoC

As the complexity of evolving integration circuits and number of cores in each chip increases, the performance aspect becomes an important issue while designing complex chips. Networking becomes difficult in such systems when compared to conventional systems. A special networking technique called Network on Chip is used in SoC to address the above problem. In this paper, two techniques such as Platform level design and Modified Random Arbiter combined with deterministic XY routing algorithm are proposed to improve the performance of NoC. For improving the NoC performance, the Platform level design technique is used for forwarding the error free packets to the cluster level network. After the error free packets are received by the router, its random arbiter will service all the incoming packets without stacking in a faster manner. This combined technique optimizes the packet storage area and also improves the performance of NoC by avoiding packet collision, deadlock and live lock conditions.

Novel Zero Crosstalk Encoding/decoding techniques for SOC

Todays high density fabrications on chip have resulted in closer interconnects. As a downside, external radiation, crosstalk coupling, supply voltage fluctuations, temperature variations, electromagnetic interference (EMI) or combinations of these manufacturing defects have increased. Of these the major concern is that of crosstalk and its consequences. In the past, multiple techniques such as shielding, use of repeaters and guard rings were proposed to alleviate the undesired signal transitions. These bygone techniques where debunked since implementing them resulted in additional wires which increases routing area. These shortcomings laid the path towards innovative encoding schemes like the No Adjacent Transitions (NAT), Boundary Shift Coding (BSC) and Bus invert technique etc. This paper proposes a novel Zero Crosstalk Encoding Scheme and explores its potential applications to alleviate the concerns of crosstalk in decisive communication modules. The Zero Crosstalk Scheme generates codes which results in a 100% reduction in unenviable signaling transitions for Type-3 and Type-4 crosstalk.

Speech enhancement using temporal masking in presence of near-end noise

In mobile phones, perceived speech signals deteriorates significantly in the presence of background noise as it arrives at the near-end listeners ears. There is an inherent need to improve the intelligibility and quality of the received speech signal in noisy environments by developing speech enhancement algorithms. This paper focuses on the speech enhancement method by incorporating temporal masking properties of the human ear. The gain adjustment process employing both pre- and post-masking are adopted to improve the intelligibility and quality of the speech signal This is implemented by automatically enhancing the speech signal when the noise is dominant. The intelligibility and quality of the speech signal are measured using SII and PESQ respectively. Experimental results show improvement in the intelligibility and quality of the speech signal with the proposed method over unprocessed far-end speech signal The proposed approach is efficient in overcoming the degradation of speech signals in the presence of background noise.

Rate-reach performance improvements in ADSL interference environment using adaptive wavelet threshold

Adaptive filtering has been shown to mitigate the effect of near end crosstalk (NEXT) in an ADSL receiver. This paper discusses the cancellation of NEXT and far end crosstalk (FEXT) using a non-parametric wavelet denoising technique. Crosstalk noise is estimated using heuristic Steinpsilas unbiased risk estimate (SURE) and Bayesian risk minimization shrinkage function. The noise level estimated by a shrinkage function is used by thresholding rules to mitigate the effect of noise in the signal. A comparison of rates achieved using these shrinkage functions with three thresholding techniques viz; soft thresholds, hard thresholds and a garrote threshold have been presented. Results obtained with AWGN, 24 ADSL NEXT and 24 ADSL FEXT disturbers show a significant improvement in Rate reach performance upwards of 4 Mbps. The computational complexity in the wavelet denoising technique is independent of the number of crosstalk sources and hence can be suitable candidate for a practical deployment of mitigation of cross talk.