Ear sensor assembly for speech processing
An ear sensor apparatus includes a microphone positionable with respect to an ear canal to detect sound signals from the ear canal and to generate a corresponding output signal. The apparatus may include a seal for substantially blocking ambient sound signals from reaching the microphone and an interface to convey the output signal to a remote speech recognition apparatus. The output signal may be normalized so that the normalized output signal is more representative of speech from a user's mouth than speech detected from the user's ear.
1. A sound processing apparatus, comprising:
a microphone positionable with respect to an ear canal to detect sound signals from the ear canal and to generate a corresponding output signal, wherein the output signal from the microphone corresponds to a portion of speech and has a different spectral response than a spectral response of the same speech from the mouth of a user; and
a circuit that normalizes the output signal from the microphone by applying a normalization function to the output signal from the microphone, the normalized signal having a spectral response that is closer to the spectral response of the speech from the mouth of the user than the spectral response of the output signal from the microphone;
wherein the normalization function is an inverse of a transfer function that is based on a ratio of a spectral response of a speech sample detected from the mouth of a sampled individual to a spectral response of the speech sample detected from the ear of the sampled individual.
2. The sound processing apparatus of claim 1 , wherein the microphone and the circuit are contained within a housing that is retained by the ear of the user.
3. The sound processing apparatus of claim 1 , wherein the microphone is contained within an ear piece that is retained by the ear of the user and the circuit forms part of a separate device that is interfaced with the ear piece.
4. The sound processing apparatus of claim 1 , wherein the transfer function is a composite function that is combination of plural spectral response ratios taken from plural individuals.
5. The sound processing apparatus of claim 1 , wherein the transfer function is a combination of a series of ratio values of the spectral response of the speech detected from the ear to the spectral response of the speech detected from the mouth for each of plural time frames of the speech sample.
6. The sound processing apparatus of claim 5 , wherein the time frames are about 10 milliseconds to about 30 milliseconds long.
7. The sound processing apparatus of claim 1 , wherein the transfer function is user dependent and the speech sample is generated by the user of the sound processing apparatus.
8. The sound processing apparatus of claim 1 , wherein the transfer function is user independent and the speech sample is generated by at least one speaker different than the user of the sound processing apparatus.
9. A sound processing apparatus, comprising:
a microphone positionable with respect to an ear canal to detect sound signals from the ear canal and to generate a corresponding output signal, wherein the output signal from the microphone corresponds to a portion of speech and has a different spectral response than a spectral response of the same speech from the mouth of a user; and
a circuit that normalizes the output signal from the microphone by applying a normalization function to the output signal from the microphone, the normalized signal having a spectral response that is closer to the spectral response of the speech from the mouth of the user than the spectral response of the output signal from the microphone;
wherein the normalization function is an inverse of a transfer function that is based on a ratio of a time domain response of a speech sample detected from the mouth of a sampled individual to a time domain response of the speech sample detected from the ear of the sampled individual.
10. A sound processing apparatus, comprising:
a microphone positionable with respect to an ear canal to detect sound signals from the ear canal and to generate a corresponding output signal, wherein the output signal from the microphone corresponds to a portion of speech and has a different spectral response than a spectral response of the same speech from the mouth of a user; and
a circuit that normalizes the output signal from the microphone by applying a normalization function to the output signal from the microphone, the normalized signal having a spectral response that is closer to the spectral response of the speech from the mouth of the user than the spectral response of the output signal from the microphone;
wherein the normalization function is adaptively established by an adaptive filter that models a difference between a sound signal from the ear and a corresponding sound signal from the mouth;
wherein the normalization function is based on a transformation matrix that is selected to decorrelate a current value of the output signal from the microphone from a past history of output signal values.
11. A sound processing apparatus, comprising:
a microphone positionable with respect to an ear canal to detect sound signals from the ear canal and to generate a corresponding output signal, wherein the output signal from the microphone corresponds to a portion of speech and has a different spectral response than a spectral response of the same speech from the mouth of a user; and
a circuit that normalizes the output signal from the microphone by applying a normalization function to the output signal from the microphone, the normalized signal having a spectral response that is closer to the spectral response of the speech from the mouth of the user than the spectral response of the output signal from the microphone;
wherein the normalization function is adaptively established by an adaptive filter that models a difference between a sound signal from the ear and a corresponding sound signal from the mouth;
wherein the normalization function is based on a transformed version of an input vector that models the difference between the sound signal from the ear and the corresponding sound signal from the mouth.
12. A sound processing apparatus, comprising:
a microphone positionable with respect to an ear canal to detect sound signals from the ear canal and to generate a corresponding output signal, wherein the output signal from the microphone corresponds to a portion of speech and has a different spectral response than a spectral response of the same speech from the mouth of a user; and
a circuit that normalizes the output signal from the microphone by applying a normalization function to the output signal from the microphone, the normalized signal having a spectral response that is closer to the spectral response of the speech from the mouth of the user than the spectral response of the output signal from the microphone;
wherein the normalization function is adaptively established by an adaptive filter that models a difference between a sound signal from the ear and a corresponding sound signal from the mouth;
wherein the normalization function is based on a non-transformed version of an input vector that models the difference between the sound signal from the ear and the corresponding sound signal from the mouth.
13. A sound processing apparatus, comprising:
a microphone positionable with respect to an ear canal to detect sound signals from the ear canal and to generate a corresponding output signal, wherein the output signal from the microphone corresponds to a portion of speech and has a different spectral response than a spectral response of the same speech from the mouth of a user; and
a circuit that normalizes the output signal from the microphone by applying a normalization function to the output signal from the microphone, the normalized signal having a spectral response that is closer to the spectral response of the speech from the mouth of the user than the spectral response of the output signal from the microphone;
wherein the normalization function is adaptively established by an adaptive filter that models a difference between a sound signal from the ear and a corresponding sound signal from the mouth;
wherein the normalization function is established by sampling speech data from an ear and a mouth of a user for a period of time and adaptively generating the normalization function over the period of time.
14. The sound processing apparatus of claim 13 , wherein the adaptive updating of the normalization function is stopped after the period of time.
15. The sound processing apparatus of claim 1 , further comprising a host that executes speech recognition software to process the normalized output signal.
16. The sound processing apparatus of claim 1 , further comprising a communications device that transmits the normalized output signal to a remote destination.