Methods and apparatus for dynamic low frequency noise suppression
View Patent ↗Methods and apparatus for dynamically suppressing low frequency non-speech audio events, such as road bumps, without suppressing speech formants. In exemplary embodiments of the invention, maximum powers in first and second windows are computed and used to determine whether dampening should be applied, and if so, to what extent.
1. A method for speech signal enhancement by dynamically suppressing low frequency noise events without suppressing speech components, comprising:
receiving an input signal;
forming a first window of the input signal spanning a first frequency range corresponding to a fundamental frequency of human voiced speech for capturing a speech formant;
forming a second window of the input signal having a second frequency range adjacent to the first frequency range;
determining information on any signal peaks in the first and second windows;
computing, using a computer processor, a dampening level from the information on the signal peaks in the first and second windows;
increasing the dampening level of the first frequency range when a harmonic of the speech formant in the first window is not detected in the second window based upon the signal peak information in the first and second windows;
adjusting sizes of the first and second windows until a final dampening level is determined for dynamically suppressing non-speech audio events in the input signal; and
outputting the input signal having the final dampening level for a loudspeaker to generate sound.
2. The method according to claim 1 , wherein the information on the signal peaks comprises a maximum power.
3. The method according to claim 2 , wherein the dampening level is computed using a ratio of the maximum powers in the first and second windows.
4. The method according to claim 1 , wherein the final dampening level corresponds to a total dampening for the first window that is maximized.
5. The method according to claim 1 , further including adjusting the sizes of the first and second windows by increasing a size of the first window and increasing a size of the second window, wherein the adjusted first and second windows do not overlap and remain adjacent to each other.
6. The method according to claim 1 , wherein the final dampening level is only applied to the first window.
7. The method according to claim 1 , wherein the first and second windows are of equal size.
8. The method according to claim 1 , further including providing a background noise floor.
9. The method according to claim 1 , wherein the first frequency range has a maximum corresponding to maximum frequency for a lowest expected speech formant.
10. The method according to claim 1 , wherein the non-speech audio event comprises a road bump.
11. The method according to claim 1 , further including making a frame-by-frame voiced/unvoiced determination and selecting a maximum frequency for the first frequency range based upon the determination of whether speech is present.
12. The method according to claim 1 , further including limiting a maximum frequency of the second frequency range based upon a maximum fundamental frequency for speech.
13. A system for speech signal enhancement by dynamically suppressing low frequency noise events without suppressing speech components, comprising:
a dynamic noise suppression module, comprising:
a frame module to sample an input signal;
a window generation module coupled to the frame module to form a first window spanning a first frequency range and a second window having a second frequency range adjacent to the first frequency range and to adjust the first and second windows, wherein the first window corresponds to a fundamental frequency of human voiced speech for capturing a speech formant;
a power module to determine signal peak information for the first window and for the second window; and
a dampening computation module to compute a dampening level corresponding to the signal peak information in the first and second windows for suppressing non-speech audio events in the input signal including increasing the dampening level of the first frequency range when a harmonic of the speech formant in the first window is not detected in the second window based upon the signal peak information in the first and second windows and to output the input signal having the final dampening level for a loudspeaker to generate sound.
14. The system according to claim 13 , wherein the dampening computation module can compute the dampening level using a ratio of the maximum powers in the first and second windows.
15. The system according to claim 13 , wherein the a window generation module can adjust the sizes of the first and second windows by increasing a size of the first frequency range and increasing a size of the second window, wherein the adjusted first and second windows do not overlap and remain adjacent to each other.
16. An article comprising:
a non-transitory computer readable medium including stored instructions that enable a machine to:
receive an input signal;
form a first window spanning a first frequency range corresponding to a fundamental frequency of human voiced speech for capturing a speech formant;
form a second window having a second frequency range adjacent to the first frequency range;
determine information on any signal peaks in the first and second windows;
compute, using a computer processor, a dampening level from the information on the signal peaks in the first and second windows;
increase the dampening level of the first frequency range when a harmonic of the speech formant in the first window is not detected in the second window based upon the signal peak information in the first and second windows;
adjust sizes of the first and second windows until a final dampening level is determined for suppressing non-speech audio events in the input signal; and
output the input signal having the final dampening level for a loudspeaker to generate sound.
17. The article according to claim 16 , further including instructions for computing the dampening level using a ratio of maximum powers in the first and second windows.
18. The article according to claim 16 , further including instructions for adjusting the sizes of the first and second windows by increasing a size of the first frequency range and increasing a size of the second window, wherein the adjusted first and second windows do not overlap and remain adjacent to each other.