PITCH DETECTION ALGORITHM BASED ON PWVT

• US 20170287507A1
• Filed: 03/31/2017
• Published: 10/05/2017
• Est. Priority Date: 03/31/2016
• Status: Active Grant

• Current Assignee: Omnispeech LLC
• Original Assignee: Omnispeech LLC
• Inventors: KHANAGHA, Vahid, KEEL, Alton
• Granted Patent: US 10,510,363 B2
• Time in Patent Office: Days
• CPC Class Codes: G10L 15/1815 Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning
• CPC Class Codes: G10L 2015/223 Execution procedure of a spoken command
• CPC Class Codes: G10L 25/06 the extracted parameters being correlation coefficients
• CPC Class Codes: G10L 25/18 the extracted parameters being spectral information of each sub-band
• CPC Class Codes: G10L 25/90 Pitch determination of speech signals

Abstract

A pitch detection method. Such a pitch detection method may apply Pseudo Weigner Ville Transformation (PWVT) as a spectral representation of speech signal. Also, the pitch detection method may take the median value of each frame of the speech signal as a threshold for making the voicing decision. Additionally, the pitch detection method may take a moving average of PWVT as the threshold for voicing decision.

Claims

1. A method of a pitch detection comprising:

sampling a signal to generate a discrete time signal;

generating a complex valued discrete time signal from the discrete time signal;

computing Pseudo Weigner Ville Transformation (PWVT) on the complex valued discrete time signal to generate a spectral representation of the signal;

computing a harmonic summation on the spectral representation to generate a pitch candidate; and

deciding the pitch candidate as a pitch value if a harmonic summation value of the pitch candidate is larger than a threshold.

2. The method of claim 1, further comprising:

applying Hilbert Transform to the discrete time signal to generate the complex valued discrete time signal, wherein the complex valued discrete time signal is low-pass filtered, buffered and down-sampled.

3. The method of claim 2, wherein when applying Hilbert Transform, a half band Finite-Impulse-Response (FIR) filter is applied to the discrete time signal for a real part of the complex valued discrete time signal, and the discrete time signal is shifted in an amount of π

/2 for an imaginary part of the complex valued discrete time signal.

4. The method of claim 2, wherein an anti-aliasing low-pass filter with cut-off frequency of 1 kHz is applied to low-pass filter the complex valued discrete time signal.

5. The method of claim 2, wherein the complex valued discrete time signal is buffered to form the analysis window with a 60 msec length.

6. The method of claim 2, wherein the complex valued discrete time signal is down sampled by a factor 4.

7. The method of claim 1, wherein the PWVT is computed by Fourier Transform as a frequency domain representation of the signal.

8. The method of claim 1, wherein the PWVT is computed by 512 points Discrete Fast Fourier Transform (DFFT) of a down-sampled Hilbert transform of the signal to form a frequency domain representation of the signal over time in the frequency range of 0 to 1 Khz.

9. The method of claim 1, wherein the harmonic summation is computed within a human pitch frequency range.

10. The method of claim 1, wherein the pitch candidate is a corresponding frequency location of a maximum value of the harmonic summation.

11. The method of claim 1, wherein the threshold is fifty times of a median value of the spectral representation.

12. The method of claim 1, wherein the signal is each frame of a speech signal.

14. The method of claim 12, wherein the threshold comprises a first threshold and a second threshold, the first threshold is the median value of the spectral representation which is generated by the PWVT and the second threshold is an average value of maximum values of harmonic summations of the frames of the speech signal in which the pitch value is not detected, and the average value is computed over more than one hundred frames of the speech signal in which the pitch value is not detected.