초록 ▼

An echo canceller for improved recognition and removal of an echo from a communication device. The echo canceller can dynamically reduce echo using an improved energy estimator and an improved adaptive filter. The improved energy estimator can determine if conversation is in a single talk period or

An echo canceller for improved recognition and removal of an echo from a communication device. The echo canceller can dynamically reduce echo using an improved energy estimator and an improved adaptive filter. The improved energy estimator can determine if conversation is in a single talk period or a double talk period based on the combined energy of both the near end background noise and speech. The improved adaptive filter can reduce echo by dynamically changing adaptation speed or step size. In double talk, the adaptive filter(s) can dynamically slow-down or stop adaptation. In single talk, the filter can dynamically increase the speed of adaptation to improve accuracy, or decrease adaptation speed for stability.

대표청구항 ▼

1. A communication device comprising: an input configured to receive a signal comprising near end background noise, near end speech signals, and far end echo signals;an estimator module configured to estimate an energy level of the near end background noise and a level of the near end speech signals

1. A communication device comprising: an input configured to receive a signal comprising near end background noise, near end speech signals, and far end echo signals;an estimator module configured to estimate an energy level of the near end background noise and a level of the near end speech signals to determine whether the communication device is in a double talk mode; andan adaptive filter configured to reduce the far end echo signals using a variable step size algorithm, wherein: the variable step size algorithm dynamically changes a speed of the adaptive filter;the variable step size algorithm dynamically decreases the step size of the adaptive filter to a non-zero amount to slow down adaptation of the adaptive filter when the communication device is in the double talk mode; andupon a determination that the adaptive filter has converged to a true room response when the communication device is determined to not be in the double talk mode, the variable step size algorithm decreases the step size of the adaptive filter to slow down adaptation and stay close to the true room response. 2. The communication device of claim 1, wherein the speed of the adaptive filter is dynamically decreased by dynamically decreasing a step size or by dynamically increasing a regularization parameter. 3. The communication device of claim 1, wherein the variable step size algorithm dynamically decreases the speed of the adaptive filter when the communication device is in a double talk period. 4. The communication device of claim 1, wherein the variable step size algorithm dynamically increases the speed of the adaptive filter when at least one of following occurs: the signal contains only far end echo signals or the adaptive filter is away from the true room response. 5. The communication device of claim 1, wherein the variable step size algorithm dynamically changes a speed of the adaptive filter during echo path changes. 6. The communication device of claim 1, wherein the adaptive filter solves an adaptation control problem during single talk periods and solves the double talk problem during double talk periods. 7. The communication device of claim 1, wherein the adaptive filter reduces the far end echo signals without receiving inputs from a double talk detection algorithm external to the adaptive filter. 8. A communication device comprising: an input configured to receive a signal comprising near end background noise, near end speech signals, and far end echo signals;an estimator module configured to estimate an energy level of the near end background noise and a level of the near end speech signals to determine whether the communication device is in a double talk mode; andan adaptive filter configured to reduce the far end echo signals using a variable step size algorithm, wherein the variable step size algorithm dynamically changes a speed of the adaptive filter, andwherein the estimator module is configured to estimate an energy level of the near end background noise and a level of the near end speech signals using the equation: γ^2⁡[n]=σ^e2⁡[n]-1l⁢σ^x2⁡[n]⁢r^exT⁡[n]⁢r^ex⁡[n],where:γ2[n] is an estimate of a near end signal energy estimator value,{circumflex over (σ)}e2[n] is an estimated variance of a residual error signal,{circumflex over (σ)}x2[n] is an estimated variance of an excitation signal vector,{circumflex over (r)}ex[n] is an estimated cross-correlation between a far end signal vector and a residual error signal, and{circumflex over (r)}Tex is a transpose vector of {circumflex over (r)}ex[n]. 9. A communication device comprising: an input configured to receive a signal comprising near end background noise, near end speech signals, and far end echo signals;an estimator module configured to estimate an energy level of the near end background noise and a level of the near end speech signals to determine whether the communication device is in a double talk mode; andan adaptive filter configured to reduce the far end echo signals using a variable step size algorithm, wherein the variable step size algorithm dynamically changes a speed of the adaptive filter, andwherein the adaptive filter is configured to reduce the far end echo signals using a variable step size algorithm to determine a variable step size using the equation: μ⁡[n]=μNEW⁡[n]xT⁡[n]⁢x⁡[n],where:μ[n] is a classical normalized least mean square solution for step size,μNEW[n] is the variable step size,x[n] is a far end signal vector, andxT[n] is a transpose vector of x[n]. 10. The communication device of claim 1, wherein the adaptive filter uses a least mean square (LMS) based methods or variants of LMS. 11. The communications device of claim 1, wherein the estimator module and the adaptive filter operate in the frequency domain or the time domain. 12. The communication device of claim 11, wherein the speed of the adaptive filter is dynamically decreased by dynamically decreasing a step size or by dynamically increasing a regularization parameter. 13. The communication device of claim 11, wherein the variable step size algorithm dynamically decreases the speed of the adaptive filter when at least one of following occurs: the communication device is in a double talk period or the adaptive filter is near the true room response. 14. The communication device of claim 11, wherein the variable step size algorithm dynamically increases the speed of the adaptive filter when at least one of following occurs: the signal contains only far end echo signals or the adaptive filter is away from the true room response. 15. The communication device of claim 11, wherein the variable step size algorithm dynamically changes a speed of the adaptive filter during echo path changes. 16. The communication device of claim 11, wherein the adaptive filter solves an adaptation control problem during single talk periods and solves the double talk problem during double talk periods. 17. The communication device of claim 11, wherein the adaptive filter filters the far end echo signals without receiving inputs from a double talk detection algorithm external to the adaptive filter. 18. A communication device comprising: an input configured to receive a signal comprising near end background noise, near end speech signals, and far end echo signals;an estimator module configured to estimate an energy level of the near end background noise and a level of the near end speech signals to determine whether the communication device is in a double talk mode; andan adaptive filter configured to reduce the far end echo signals using a variable step size algorithm,wherein: the variable step size algorithm dynamically changes a speed of the adaptive filter;the estimator module and the adaptive filter operate in the frequency domain or the time domain; andthe estimator module is configured to estimate an energy level of the near end background noise and a level of the near end speech signals using the equation: γ^2⁡[n]=σ^e2⁡[n]-1l⁢σ^x2⁡[n]⁢r^exT⁡[n]⁢r^ex⁡[n],where:γ2[n] is an estimate of a near end signal energy estimator value,{circumflex over (σ)}e2[n] is an estimated variance of a residual error signal,{circumflex over (σ)}x2[n] is an estimated variance of an excitation signal vector,{circumflex over (r)}ex[n] is an estimated cross-correlation between a far end signal vector and a residual error signal, and{circumflex over (r)}Tex[n] is a transpose vector of {circumflex over (r)}ex[n]. 19. A communication device comprising: an input configured to receive a signal comprising near end background noise, near end speech signals, and far end echo signals;an estimator module configured to estimate an energy level of the near end background noise and a level of the near end speech signals to determine whether the communication device is in a double talk mode; andan adaptive filter configured to reduce the far end echo signals using a variable step size algorithm, wherein the variable step size algorithm dynamically changes a speed of the adaptive filter,wherein the estimator module and the adaptive filter operate in the frequency domain or the time domain, and wherein the adaptive filter is configured to reduce the far end echo signals using a variable step size algorithm to determine a variable step size using the equation: μ⁡[n]=μNEW⁡[n]xT⁡[n]⁢x⁡[n],where:μ[n] is a classical normalized least mean square solution for step size,μNEW[n] is the variable step size,x[n] is a far end signal vector, andxT[n] is a transpose vector of x[n]. 20. The communication device of claim 11, wherein the adaptive filter uses a least mean square (LMS) based methods or variants of LMS. 21. The communication device of claim 1, wherein the communication device comprises one of a cellular telephone or a bluetooth communication device. 22. The communication device of claim 1, wherein the input comprises at least one microphone. 23. A method of reducing echo signals during wireless communication, comprising: receiving a voice signal comprising near end background noise, near end speech signals, and far end echo signals;estimating an energy level comprising both the near end background noise and the near end speech signals;determining whether said wireless communication is in a double talk period based on the estimated energy level; andreducing said echo signals using a variable step size adaptive filter, wherein reducing said echo signals comprises decreasing the step size of the adaptive filter upon a determination that the adaptive filter has converged to a true room response when the wireless communication is determined to not be in the double talk period, wherein decreasing step size slows down adaptation of the adaptive filter and allows the adaptive filter to stay close to the true room response, and wherein reducing said echo signals also includes dynamically decreasing the step size of the adaptive filter to a non-zero amount to slow down, without initially stopping, adaptation of the adaptive filter when the communication device is in the double talk period. 24. The method of claim 23, wherein reducing said echo signals further comprises dynamically reducing the speed of the adaptive filter when there is double talk. 25. The method of claim 23, wherein reducing said echo signals further comprises dynamically decreasing a step size or dynamically increasing a regularization parameter when there is double talk. 26. The method of claim 23, further comprising increasing the speed of the adaptive filter when at least one of following occurs: said wireless communication is not in a double talk period or the adaptive filter estimate is away from the true room response. 27. The method of claim 23, wherein reducing said echo signals comprises filtering the echo signals without inputs from a double talk detection algorithm external to the adaptive filter. 28. An echo cancellation module comprising: an estimator module configured to estimate the energy of a near end sound signal; andan adaptive filter configured to: determine when said sound signal is in a double talk period;reduce a far end echo in response to said signal being in a double talk period, wherein the adaptive filter adjusts a variable step size based on the estimated energy;dynamically decreases the step size of the adaptive filter to a non-zero amount to slow down adaptation of the adaptive filter when the communication device is in the double talk period; anddecrease the step size upon a determination that the adaptive filter has converged to a true room response when the sound signal is determined to not be in the double talk period, wherein decreasing the step size slows down adaptive filter adaptation and allows the adaptive filter to stay close to the true room response. 29. The echo cancellation module of claim 28, wherein the adjustment of the adaptive filter comprises dynamically reducing the variable step size to decrease a speed of the adaptive filter. 30. The echo cancellation module of claim 29, wherein the variable step size is dynamically reduced during double talk. 31. The echo cancellation module of claim 28, wherein the adjustment of the adaptive filter comprises dynamically increasing the variable step size to increase a speed of the adaptive filter. 32. The echo cancellation module of claim 28, wherein the adaptive filter provides filtering without receiving inputs from a double talk detection algorithm external to the filter. 33. A wireless communication device, comprising: means for receiving a voice signal comprising near end background noise, near end speech signals, and far end echo signals;means for estimating an energy level comprising both the near end background noise and the near end speech signals;means for determining whether said wireless communication is in a double talk period based on the estimated energy level;means for reducing said echo signals using a variable step size adaptive filter;means for dynamically decreasing the step size of the variable step size adaptive filter to a non-zero amount to slow down adaptation of the variable step size adaptive filter when the communication device is in the double talk period; andmeans for decreasing the step size of the adaptive filter upon a determination that the adaptive filter has converged to a true room response when said wireless communication is determined to not be in the double talk period, wherein decreasing the step size slows down the adaptive filter adaptation and allows the adaptive filter to stay close to the true room response. 34. The wireless communication device of claim 33, wherein the means for reducing comprises means for dynamically decreasing a speed by dynamically decreasing a step size or by dynamically increasing a regularization parameter. 35. The wireless communication device of claim 33, wherein the means for reducing solves an adaptation control problem during single talk periods and solves a double talk problem during double talk periods. 36. A non-transitory computer-readable storage medium comprising instructions that, when executed by a computer, cause the computer to perform the method of: receiving a voice signal comprising near end background noise, near end speech signals, and far end echo signals;estimating an energy level comprising both the near end background noise and the near end speech signals;determining whether said wireless communication is in a double talk period based on the estimated energy level; andreducing said echo signals using a variable step size adaptive filter, wherein reducing said echo signals comprises: dynamically decreasing the step size of the adaptive filter to a non-zero amount to slow down adaptation of the adaptive filter when the communication device is in the double talk period;decreasing the step size of the adaptive filter upon a determination that the adaptive filter has converged to a true room response when the wireless communication is determined to not be in a double talk period, wherein reducing step size slows down adaptation of the adaptive filter and allows the adaptive filter to stay close to the true room response. 37. The non-transitory computer-readable storage medium of claim 36, further comprising instructions for dynamically increasing a speed of the variable step size adaptive filter: during single talk periods; orwhen the adaptive filter is away from the true room response. 38. The non-transitory computer-readable storage medium of claim 36, wherein the adaptive filter does not receive inputs from a double talk detection algorithm external to the adaptive filter.