There is a great need for speech enhancement in today's world due to the increasing demand for sp... more There is a great need for speech enhancement in today's world due to the increasing demand for speech based applications. These applications vary from hearing-aids, hands-free telephony to speech controlled devices. The main goal is to minimize the interference from an acquired speech signal. The interference we considered here could be from any noise source such as competing speaker, radio, TV and so on. This paper proposes a solution to improve the current design of the switched Griffiths-Jim beamformer structure. It introduces an adaptive nonlinear neural network algorithm for the noise reduction section. The network topology used here is a partially connected three-layer feedforward neural network structure. The error backpropagation algorithm is used here as the learning algorithm. Comparison analysis of the traditional four channel linear beamformer and the proposed four-channel neural switched Griffiths-Jim beamformer structure is discussed here. They are both tested with different types of interference signal from the Noise-X database. All the experiments are conducted in real-world surrounding. The nonlinear approach introduced here shows remarkable improvement over the previous linear adaptive beamformer approach.
This paper proposes a novel multi-microphone nonlinear neural network based switched Griffiths-Ji... more This paper proposes a novel multi-microphone nonlinear neural network based switched Griffiths-Jim beamformer structure for speech enhancement. The main objective of this algorithm is to reduce real-world interference signals such as radio, television or computer fan noise from an acquired speech signal. The proposed algorithm improves the current design of the switched Griffiths-Jim beamformer structure by introducing an adaptive nonlinear neural network filter for the noise reduction section. The network topology used here is a partially connected three-layer feedforward neural network structure. The error backpropagation algorithm is used here as the learning algorithm. A comparison analysis of the traditional three-microphone linear beamformer and the proposed three-microphone neural switched Griffiths-Jim beamformer structure is discussed here. They are both tested with different types of interference signal from the Noise-X database. All the experiments are conducted in real-world surroundings. The nonlinear approach introduced here shows remarkable improvement over the previous linear adaptive beamformer approach.
This paper presents a special nonlinear switched Griffiths-Jim beamformer (SGJBF) structure. The ... more This paper presents a special nonlinear switched Griffiths-Jim beamformer (SGJBF) structure. The main objective of this paper is to reduce the background noise from an acquired speech signal. The interference we considered here is non-stationary in nature and can arrive from a variety of potential sources; for example, competing talkers, radio, TV and so on. In this paper, we propose an adaptive Time Delay Neural Network (TDNN) based nonlinear noise canceller. The proposed structure consists of a three-layer feedforward network with partially connected layers to achieve real-time processing. The error backpropagation learning algorithm is used here to train the TDNN. This system is tested with different types of interference signals from the Noise-X database. A comparison analysis of the proposed structure and the traditional linear adaptive beamformer is presented here. The nonlinear approach investigated here show remarkable improvements over the previous linear based beamforming approach.
Speech controlled applications are now becoming more and more practical due to advances in techno... more Speech controlled applications are now becoming more and more practical due to advances in technology. These applications vary from command and control instruments, video conferencing to robotics. However, their performance decreases when the acquired speech signal is corrupted by background noise. Numerous research has been done in the last two decades to improve their performance. The switched Griffiths-Jim beamformer is one of the well known methods used to reduce background noise (or interference). This algorithm makes use of two adaptive filters and a voice activity detector. The first filter is used as a beam-steering filter which is updated during speech signals and the second filter is used as a noise cancellation filter which is updated during a noise-alone signal. A voice activity detector is used to control the updates of these adaptive filters. Generally, the normalised least mean squares (NLMS) algorithm is used for the adaptive filters. However, it is found that the convergence rate of the NLMS deteriorates for coloured noise under certain non-stationary conditions where the correlation matrix is "stiff". It has been found that the affine projection algorithm will perform better under these conditions. Therefore, a comparison of these two adaptive filters in this beamformer structure will be discussed in this paper.
In recent years, speech acquisition in an adverse environment has received considerable attention... more In recent years, speech acquisition in an adverse environment has received considerable attention due to the increased need for hands-free and voice controlled applications. The main problem with acquiring a speech signal from the environment is that it is often corrupted by background noise. This background noise could be from several components propagating from different sources such as radio, TV, a computer fan, or another talker. For the voice controlled applications to work efficiently they require a high quality speech signal. It is therefore necessary to reduce the interferences from the received signal to obtain the actual speech signal.
In some applications headsets or body worn microphones are used to get around this problem. However, using a close talking microphone is impractical in some places. Directive microphones can be used to get around this problem, but they don’t reach their objective in adverse environments. Another potential solution to this problem is to use a microphone array. The microphone array makes use of beamforming techniques to fight against the effects of the acoustic environment.
The beamforming technique used here is the well-known modified version of the Griffiths-Jim beamformer. Van Compernolle and Leuven discovered this algorithm in 1990. This algorithm makes use of two adaptive filters based on Least Mean Squares (LMS). Since the LMS algorithm has some drawbacks with stability and selection of the step-size, we will be using instead an adaptive filter based on Normalized LMS (NLMS).
This first NLMS algorithm updates during a speech signal and, the second NLMS algorithm updates during the noise signal. The first one acts as a beam-steering filter, and the second one acts as a filter for the noise. Only one of the NLMS algorithms is updated at a given time. The technique also uses a simple Voice Activity Detector (VAD) to analyse the received speech signal and determine if it is speech or noise. The corresponding NLMS algorithm is updated depending on the result obtained from this VAD.
Much of the work in the smart house technology has been done on individual technologies, but litt... more Much of the work in the smart house technology has been done on individual technologies, but little has been done on their integration into a cohesive whole. The Bluetooth house project at Massey University in New Zealand, which was initiated in 2002, embraced a systems engineering approach to design a usable smart house, aiming at a complete and integrated solution, which can be customised, based on individual needs, to give elderly people independence, quality of life, and the safety they require. This paper presents how the Massey Bluetooth smart house design project has been carried out and what the smart home may look like in the near future. Considering current technical feasibility and the advances in other research, it is suggested that for a house to be considered as truly ‘smart’, four levels of smartness are imperative: smart sensors, smart management, smart control, and smart appliances. The Bluetooth house at Massey University incorporates these four smart technologies and allows all these individual technologies to be integrated into a seamless whole. For smart sensing, the project employed Bluetooth technology to connect the whole house, and to locate the user’s position. In order to coordinate all the technologies, a smart management system was developed, that is capable of coordinating the information for commands, feedback from smart appliances, and user’s location information. It can make intelligent decisions on what to do, or relay necessary information to individual intelligent devices throughout the house. In addition, the medium of communication with the house must be as natural as possible, in order to make it as easy as possible for the occupants of the smart house to interact with and the various smart appliances. A voice-activated universal remote control and a new microphone system are being developed to this end. Finally, the smart house has to provide an enjoyable experience that can promote the uptake of smart house technology by users in the future. An interactive TV environment is being developed to this end. The Massey Bluetooth house project is not so much aimed at a cutting-edge technology in smart house design, but at integrating technologies into a seamless, cohesive whole through the application of four levels of smartness.
There is a great need for speech enhancement in today's world due to the increasing demand for sp... more There is a great need for speech enhancement in today's world due to the increasing demand for speech based applications. These applications vary from hearing-aids, hands-free telephony to speech controlled devices. The main goal is to minimize the interference from an acquired speech signal. The interference we considered here could be from any noise source such as competing speaker, radio, TV and so on. This paper proposes a solution to improve the current design of the switched Griffiths-Jim beamformer structure. It introduces an adaptive nonlinear neural network algorithm for the noise reduction section. The network topology used here is a partially connected three-layer feedforward neural network structure. The error backpropagation algorithm is used here as the learning algorithm. Comparison analysis of the traditional four channel linear beamformer and the proposed four-channel neural switched Griffiths-Jim beamformer structure is discussed here. They are both tested with different types of interference signal from the Noise-X database. All the experiments are conducted in real-world surrounding. The nonlinear approach introduced here shows remarkable improvement over the previous linear adaptive beamformer approach.
This paper proposes a novel multi-microphone nonlinear neural network based switched Griffiths-Ji... more This paper proposes a novel multi-microphone nonlinear neural network based switched Griffiths-Jim beamformer structure for speech enhancement. The main objective of this algorithm is to reduce real-world interference signals such as radio, television or computer fan noise from an acquired speech signal. The proposed algorithm improves the current design of the switched Griffiths-Jim beamformer structure by introducing an adaptive nonlinear neural network filter for the noise reduction section. The network topology used here is a partially connected three-layer feedforward neural network structure. The error backpropagation algorithm is used here as the learning algorithm. A comparison analysis of the traditional three-microphone linear beamformer and the proposed three-microphone neural switched Griffiths-Jim beamformer structure is discussed here. They are both tested with different types of interference signal from the Noise-X database. All the experiments are conducted in real-world surroundings. The nonlinear approach introduced here shows remarkable improvement over the previous linear adaptive beamformer approach.
This paper presents a special nonlinear switched Griffiths-Jim beamformer (SGJBF) structure. The ... more This paper presents a special nonlinear switched Griffiths-Jim beamformer (SGJBF) structure. The main objective of this paper is to reduce the background noise from an acquired speech signal. The interference we considered here is non-stationary in nature and can arrive from a variety of potential sources; for example, competing talkers, radio, TV and so on. In this paper, we propose an adaptive Time Delay Neural Network (TDNN) based nonlinear noise canceller. The proposed structure consists of a three-layer feedforward network with partially connected layers to achieve real-time processing. The error backpropagation learning algorithm is used here to train the TDNN. This system is tested with different types of interference signals from the Noise-X database. A comparison analysis of the proposed structure and the traditional linear adaptive beamformer is presented here. The nonlinear approach investigated here show remarkable improvements over the previous linear based beamforming approach.
Speech controlled applications are now becoming more and more practical due to advances in techno... more Speech controlled applications are now becoming more and more practical due to advances in technology. These applications vary from command and control instruments, video conferencing to robotics. However, their performance decreases when the acquired speech signal is corrupted by background noise. Numerous research has been done in the last two decades to improve their performance. The switched Griffiths-Jim beamformer is one of the well known methods used to reduce background noise (or interference). This algorithm makes use of two adaptive filters and a voice activity detector. The first filter is used as a beam-steering filter which is updated during speech signals and the second filter is used as a noise cancellation filter which is updated during a noise-alone signal. A voice activity detector is used to control the updates of these adaptive filters. Generally, the normalised least mean squares (NLMS) algorithm is used for the adaptive filters. However, it is found that the convergence rate of the NLMS deteriorates for coloured noise under certain non-stationary conditions where the correlation matrix is "stiff". It has been found that the affine projection algorithm will perform better under these conditions. Therefore, a comparison of these two adaptive filters in this beamformer structure will be discussed in this paper.
In recent years, speech acquisition in an adverse environment has received considerable attention... more In recent years, speech acquisition in an adverse environment has received considerable attention due to the increased need for hands-free and voice controlled applications. The main problem with acquiring a speech signal from the environment is that it is often corrupted by background noise. This background noise could be from several components propagating from different sources such as radio, TV, a computer fan, or another talker. For the voice controlled applications to work efficiently they require a high quality speech signal. It is therefore necessary to reduce the interferences from the received signal to obtain the actual speech signal.
In some applications headsets or body worn microphones are used to get around this problem. However, using a close talking microphone is impractical in some places. Directive microphones can be used to get around this problem, but they don’t reach their objective in adverse environments. Another potential solution to this problem is to use a microphone array. The microphone array makes use of beamforming techniques to fight against the effects of the acoustic environment.
The beamforming technique used here is the well-known modified version of the Griffiths-Jim beamformer. Van Compernolle and Leuven discovered this algorithm in 1990. This algorithm makes use of two adaptive filters based on Least Mean Squares (LMS). Since the LMS algorithm has some drawbacks with stability and selection of the step-size, we will be using instead an adaptive filter based on Normalized LMS (NLMS).
This first NLMS algorithm updates during a speech signal and, the second NLMS algorithm updates during the noise signal. The first one acts as a beam-steering filter, and the second one acts as a filter for the noise. Only one of the NLMS algorithms is updated at a given time. The technique also uses a simple Voice Activity Detector (VAD) to analyse the received speech signal and determine if it is speech or noise. The corresponding NLMS algorithm is updated depending on the result obtained from this VAD.
Much of the work in the smart house technology has been done on individual technologies, but litt... more Much of the work in the smart house technology has been done on individual technologies, but little has been done on their integration into a cohesive whole. The Bluetooth house project at Massey University in New Zealand, which was initiated in 2002, embraced a systems engineering approach to design a usable smart house, aiming at a complete and integrated solution, which can be customised, based on individual needs, to give elderly people independence, quality of life, and the safety they require. This paper presents how the Massey Bluetooth smart house design project has been carried out and what the smart home may look like in the near future. Considering current technical feasibility and the advances in other research, it is suggested that for a house to be considered as truly ‘smart’, four levels of smartness are imperative: smart sensors, smart management, smart control, and smart appliances. The Bluetooth house at Massey University incorporates these four smart technologies and allows all these individual technologies to be integrated into a seamless whole. For smart sensing, the project employed Bluetooth technology to connect the whole house, and to locate the user’s position. In order to coordinate all the technologies, a smart management system was developed, that is capable of coordinating the information for commands, feedback from smart appliances, and user’s location information. It can make intelligent decisions on what to do, or relay necessary information to individual intelligent devices throughout the house. In addition, the medium of communication with the house must be as natural as possible, in order to make it as easy as possible for the occupants of the smart house to interact with and the various smart appliances. A voice-activated universal remote control and a new microphone system are being developed to this end. Finally, the smart house has to provide an enjoyable experience that can promote the uptake of smart house technology by users in the future. An interactive TV environment is being developed to this end. The Massey Bluetooth house project is not so much aimed at a cutting-edge technology in smart house design, but at integrating technologies into a seamless, cohesive whole through the application of four levels of smartness.
Uploads
Papers by Vaitheki Yoganathan
In some applications headsets or body worn microphones are used to get around this problem. However, using a close talking microphone is impractical in some places. Directive microphones can be used to get around this problem, but they don’t reach their objective in adverse environments. Another potential solution to this problem is to use a microphone array. The microphone array makes use of beamforming techniques to fight against the effects of the acoustic environment.
The beamforming technique used here is the well-known modified version of the Griffiths-Jim beamformer. Van Compernolle and Leuven discovered this algorithm in 1990. This algorithm makes use of two adaptive filters based on Least Mean Squares (LMS). Since the LMS algorithm has some drawbacks with stability and selection of the step-size, we will be using instead an adaptive filter based on Normalized LMS (NLMS).
This first NLMS algorithm updates during a speech signal and, the second NLMS algorithm updates during the noise signal. The first one acts as a beam-steering filter, and the second one acts as a filter for the noise. Only one of the NLMS algorithms is updated at a given time. The technique also uses a simple Voice Activity Detector (VAD) to analyse the received speech signal and determine if it is speech or noise. The corresponding NLMS algorithm is updated depending on the result obtained from this VAD.
In some applications headsets or body worn microphones are used to get around this problem. However, using a close talking microphone is impractical in some places. Directive microphones can be used to get around this problem, but they don’t reach their objective in adverse environments. Another potential solution to this problem is to use a microphone array. The microphone array makes use of beamforming techniques to fight against the effects of the acoustic environment.
The beamforming technique used here is the well-known modified version of the Griffiths-Jim beamformer. Van Compernolle and Leuven discovered this algorithm in 1990. This algorithm makes use of two adaptive filters based on Least Mean Squares (LMS). Since the LMS algorithm has some drawbacks with stability and selection of the step-size, we will be using instead an adaptive filter based on Normalized LMS (NLMS).
This first NLMS algorithm updates during a speech signal and, the second NLMS algorithm updates during the noise signal. The first one acts as a beam-steering filter, and the second one acts as a filter for the noise. Only one of the NLMS algorithms is updated at a given time. The technique also uses a simple Voice Activity Detector (VAD) to analyse the received speech signal and determine if it is speech or noise. The corresponding NLMS algorithm is updated depending on the result obtained from this VAD.