Regeneration of wideband speech

1. A method for regeneration of wideband speech, comprising:

receiving samples of a narrowband speech signal having a first range of frequencies, wherein a first portion of a range of frequencies in a wideband speech signal is represented in the narrowband speech signal;

identifying, based on a characteristic of the narrowband speech signal, frequencies in the first range of frequencies to translate into a target band of a regenerated speech signal, the characteristic being determined from a pitch-dependent spectral translation as approximating a harmonic structure in a second portion of the range of frequencies in the wideband speech signal, wherein the second portion of the range of frequencies is excluded from being represented in the narrowband speech signal;

modulating the identified frequencies in the first range of frequencies of the received samples of the narrowband speech signal with a modulation signal, the modulation signal having a modulating frequency adapted to upshift the identified frequencies in the first range of frequencies into the target band;

filtering the modulated samples, using a target band filter, to form the regenerated speech signal in the target band; and

combining the narrowband speech signal with the regenerated speech signal to produce a new wideband speech signal.

2. The method of claim 1 , further comprising, selecting the modulating frequency.

3. The method of claim 1 , wherein the modulating frequency matches the bandwidth of the target band.

4. The method of claim 1 , wherein the modulating frequency is normalised with respect to a sampling frequency used for generating the samples of the narrowband speech signal prior to modulation of the received samples.

5. The method of claim 1 , wherein the first range of frequencies include all the frequencies in the narrowband speech signal.

6. The method of claim 1 , wherein the signal characteristic is one of:

highest signal to noise ratio;

minimum echo;

degree of voicing; or

temporal location.

7. The method of claim 1 , wherein the target band filter is a high pass filter with a lower limit defining the lower most frequency in the target band.

8. The method of claim 1 , further comprising, controlling the filtering range of the target band filter.

9. The method of claim 1 , further comprising:

supplying the received samples of the narrowband speech signal to each of a plurality of paths;

modulating the samples on each path with a respective modulation signal;

on each path filtering the modulated samples using a high pass filter; and

combining the filtered signals to form the regenerated speech signal in the target band.

10. The method of claim 9 , further comprising:

performing low pass filtering the samples on one or more of the paths thereby to select a range of frequencies for that path.

11. The method of claim 9 , wherein the filtered signals are combined using weightings applied to each filtered signal.

12. The method of claim 1 , wherein the samples of the narrowband speech signal are received in blocks, and wherein the modulation signal includes a phase which is updated for each successive block.

13. The method of claim 1 , wherein the regenerated target band is subject to an estimated spectral envelope prior to combining the narrowband speech signal with the regenerated speech signal.

14. A computing system, comprising:

a memory, operable to host data for a narrowband speech signal, the narrowband speech signal being generated from a wideband speech signal;

a processor, operably coupled to the memory, the processor to execute instructions that cause the processor to:

receive samples of a narrowband speech signal having a first range of frequencies, wherein a first portion of a range of frequencies in a wideband speech signal is represented in the narrowband speech signal;

identify, based on a characteristic of the narrowband speech signal, frequencies in the first range of frequencies to translate into a target band of a regenerated speech signal, the characteristic being determined from a pitch-dependent spectral translation as approximating a harmonic structure in a second portion of the range of frequencies in the wideband speech signal, wherein the second portion of the range of frequencies is excluded from being represented in the narrowband speech signal;

modulate the identified frequencies in the first range of frequencies of the received samples of the narrowband speech signal with a modulation signal, the modulation signal having a modulating frequency adapted to upshift the identified frequencies in the first range of frequencies into the target band;

filter the modulated samples, using a target band filter, to form the regenerated speech signal in the target band; and

combine the narrowband speech signal with the regenerated speech signal to produce a new wideband speech signal.

15. The computing system of claim 14 , the instructions further cause the processor to:

select the modulating frequency, wherein the modulating frequency matches the bandwidth of the target band, or wherein the modulating frequency is normalised with respect to a sampling frequency used for generating the samples of the narrowband speech signal prior to modulation of the received samples.

16. The computing system of claim 14 , wherein the signal characteristic is one of:

highest signal to noise ratio;

minimum echo;

degree of voicing; or

temporal location.

17. A machine-readable storage device, excluding a transitory propagating signal, the storage device comprising instructions for execution by a processor of the machine, wherein the instructions, when executed, cause the processor to perform operations comprising:

obtaining samples of a narrowband speech signal having a first range of frequencies, wherein a first portion of a range of frequencies in a wideband speech signal is represented in the narrowband speech signal;

identifying, based on a characteristic of the narrowband speech signal, frequencies in the first range of frequencies to translate into a target band of a regenerated speech signal, the characteristic being determined from a pitch-dependent spectral translation as approximating a harmonic structure in a second portion of the range of frequencies in the wideband speech signal, wherein the second portion of the range of frequencies is excluded from being represented in the narrowband speech signal;

modulating the identified frequencies in the first range of frequencies of the obtained samples of the narrowband speech signal with a modulation signal, the modulation signal having a modulating frequency adapted to upshift the identified frequencies in the first range of frequencies into the target band;

filtering the modulated samples, using a target band filter, to form the regenerated speech signal in the target band; and

combining the narrowband speech signal with the regenerated speech signal to produce a new wideband speech signal.

18. The machine-readable storage device of claim 17 , the operations further comprising:

selecting the modulating frequency, wherein the modulating frequency matches the bandwidth of the target band, or wherein the modulating frequency is normalised with respect to a sampling frequency used for generating the samples of the narrowband speech signal prior to modulation of the obtained samples.

19. The machine-readable storage device of claim 17 , wherein the signal characteristic is one of:

highest signal to noise ratio;

minimum echo;

degree of voicing; or

temporal location.

20. The machine-readable storage device of claim 17 , wherein the first range of frequencies include all the frequencies in the narrowband speech signal.