# US 2015/0106087 A1 · Efficient Discrimination of Voiced and Unvoiced Sounds

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2015/0106087 A1
- Document ID: US20150106087A1
- Status: Published application
- Kind: A1
- Application number: 14/253,120
- Filed: 2014-04-15
- Published: 2015-04-16
- Earliest claimed: 2013-10-14
- Assignees: Zanavox
- Applicants: Zanavox
- Inventors: David Edward Newman
- Web page: https://patentgrep.com/patent/US20150106087A1

## CPC

- G10L 25/78: Detection of presence or absence of voice signals (switching of direction of transmission by voice frequency in two-way loud-speaking telephone systems H04M9/10)
- G10L: SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING

## Abstract

A method is disclosed for discriminating voiced and unvoiced sounds in speech. The method detects characteristic waveform features of voiced and unvoiced sounds, by applying integral and differential functions to the digitized sound signal in the time domain. Laboratory tests demonstrate extremely high reliability in separating voiced and unvoiced sounds. The method is very fast and computationally efficient. The method enables voice activation in resource-limited and battery-limited devices, including mobile devices, wearable devices, and embedded controllers. The method also enables reliable command identification in applications that recognize only predetermined commands. The method is suitable as a pre-processor for natural language speech interpretation, improving recognition and responsiveness. The method enables realtime coding or compression of speech according to the sound type, improving transmission efficiency.

## Claims

### Claim 1

A method for discriminating voiced sounds from unvoiced sounds in speech, comprising the steps:
converting speech sounds to a speech signal comprising sequential digitized values;
generating an integral signal by integrating the speech signal;
generating a differential signal by differentiating the speech signal;
detecting the voiced sounds by comparing the integral signal to an integral-signal threshold;
and detecting the unvoiced sounds by comparing the differential signal to a differential-signal threshold.

### Claim 2 (depends on 1)

The method of claim 1 wherein:
the speech signal includes wavelets, each wavelet comprising a monopolar variation of the speech signal with a duration in the range of 0.05 to 5 milliseconds;
the integral signal includes wavelets with durations in the range of 0.25 to 5 milliseconds and substantially excludes wavelets with durations in the range of 0.05 to 0.15 milliseconds;
and the differential signal includes wavelets with durations in the range of 0.05 to 0.15 milliseconds and substantially excludes wavelets with durations in the range of 0.25 to 5 milliseconds.

### Claim 3

A method for discriminating voiced sounds from unvoiced sounds in speech comprising the steps:
converting speech sounds to a speech signal comprising sequential digitized values;
generating an integral signal by averaging speech signal values that span a time Tinteg, wherein Tinteg is set to substantially exclude the unvoiced sounds from the integral signal;
generating a differential signal by differentiating the speech signal according to a second-order discrete differential formula that includes values that span a time Tdif, wherein Tdif is set to substantially exclude the voiced sounds from the differential signal;
detecting the voiced sounds by comparing the integral signal to an integral-signal threshold;
detecting the unvoiced sounds by comparing the differential signal to a differential-signal threshold;
and generating at least one output signal indicating when voiced and unvoiced sounds are present.

### Claim 4

A method for discriminating voiced sounds from unvoiced sounds in speech, comprising the steps:
converting speech sounds to a speech signal comprising sequential digitized values spaced apart by a time Tdata, wherein Tdata is in the range of 0.04 to 0.06 milliseconds;
generating an integral signal by averaging values of the speech signal that span a time Tinteg, wherein Tinteg is in the range of 0.2 to 0.3 milliseconds;
generating a differential signal by differentiating the speech signal according to a second-order discrete differential formula that includes three speech signal values that span a time Tdif, wherein Tdif is in the range of 0.05 to 0.15 milliseconds;
detecting the voiced sounds by comparing the integral signal to an integral-signal threshold;
and detecting the unvoiced sounds by comparing the differential signal to a differential-signal threshold.

### Claim 5 (depends on 4)

The method of claim 4 wherein the step of converting the sounds to a speech signal includes the steps:
converting sound waves to an analog electrical signal using a microphone;
converting the analog electrical signal to a digitizer-output signal comprising sequential digital values, using an analog-to-digital converter;
generating an offset signal by averaging values of the digitizer-output signal;
and generating the speech signal by subtracting the offset signal from the digitizer-output signal.

### Claim 6 (depends on 5)

The method of claim 5 which further includes the step of discarding or averaging a portion of the values of the digitizer-output signal when the temporal spacing of the digitizer-output values is less than 0.04 milliseconds.

### Claim 7 (depends on 4)

The method of claim 4 wherein the differential signal is calculated according to the formula:
D=[v(i)−2*v(i−1)+v(i−2)]/4
where the symbols v(i), v(i−1), v(i−2) represent sequential values of the speech signal and D represents the differential signal.

### Claim 8 (depends on 4)

The method of claim 4 which further includes the steps:
rectifying the integral signal, thereby producing a rectified integral signal;
rectifying the differential signal, thereby producing a rectified differential signal;
averaging the rectified integral signal with an averaging span in the range of 1 to 30 milliseconds, thereby producing a smoothed integral signal;
averaging the rectified differential signal with an averaging span in the range of 1 to 30 milliseconds, thereby producing a smoothed differential signal;
detecting the voiced sounds by comparing the smoothed integral signal to a first threshold;
and detecting the unvoiced sounds by comparing the smoothed differential signal to a second threshold.

### Claim 9 (depends on 8)

The method of claim 8 which further includes the steps:
indicating that a voiced sound has started when the smoothed integral signal exceeds an upper integral-signal threshold;
indicating that a voiced sound has ended when the smoothed integral signal drops below a lower integral-signal threshold that is lower than the upper integral-signal threshold;
indicating that an unvoiced sound has started when the smoothed differential signal exceeds an upper differential-signal threshold;
and indicating that an unvoiced sound has ended when the smoothed differential signal drops below a lower differential-signal threshold that is lower than the upper differential-signal threshold.

### Claim 10 (depends on 4)

The method of claim 4 which further includes the steps:
subtracting the integral signal from the speech signal, thereby producing a speech-minus-integral signal;
differentiating the speech-minus-integral signal, thereby producing a refined differential signal;
subtracting the differential signal from the speech signal, thereby producing a speech-minus-differential signal;
integrating the speech-minus-differential signal, thereby producing a refined integral signal;
detecting the voiced sounds by comparing the refined integral signal to a first threshold;
and detecting the unvoiced sounds by comparing the refined differential signal to a second threshold.

### Claim 11 (depends on 4)

The method of claim 4 which further includes the steps:
generating a voiced-beginning pulse when the integral signal exceeds the integral-signal threshold following a period of silence or of unvoiced sound;
and generating an unvoiced-beginning pulse when the differential signal exceeds the differential-signal threshold following a period of silence or of voiced sound.

### Claim 12 (depends on 4)

The method of claim 4 which further includes the steps:
generating a toggle signal that can be set to a first state or a second state;
setting the toggle signal to the first state when the integral signal exceeds the integral-signal threshold;
and setting the toggle signal to the second state when the differential signal exceeds the differential-signal threshold.

### Claim 13 (depends on 4)

The method of claim 4 which further includes a number Nv in the range of 2 to 50, and a number Nu in the range of 2 to 50, and the steps:
indicating that a voiced sound is present if the integral signal exceeds the integral-signal threshold Nv times while the differential signal exceeds the differential-signal threshold zero times;
and indicating that an unvoiced sound is present if the differential signal exceeds the differential-signal threshold Nu times while the integral signal exceeds the integral-signal threshold zero times.

### Claim 14 (depends on 4)

The method of claim 4 which further includes a predetermined expiration time in the range of 5 to 100 milliseconds, and the steps:
indicating that a voiced sound has started when the integral signal exceeds the integral-signal threshold;
indicating that the voiced sound has ended when the integral signal remains below the integral-signal threshold for a time equal to the expiration time;
indicating that an unvoiced sound has started when the differential signal exceeds the differential-signal threshold;
and indicating that the unvoiced sound has ended when the differential signal remains below the differential-signal threshold for a time equal to the expiration time.

### Claim 15 (depends on 4)

The method of claim 4 which further includes a voiced tally counter and an unvoiced tally counter, each tally counter comprising a counter that can be incremented and decremented, said method further including the steps:
incrementing the voiced tally counter when the integral signal exceeds the integral-signal threshold;
incrementing the unvoiced tally counter when the differential signal exceeds the differential-signal threshold;
and decrementing the voiced tally counter and the unvoiced tally counter periodically.

### Claim 16 (depends on 15)

The method of claim 15 which further includes the steps:
indicating that a voiced sound has started when the voiced tally counter exceeds a voiced-tally threshold;
indicating that the voiced sound has ended when the voiced tally counter passes below the voiced-tally threshold;
indicating that an unvoiced sound has started when the unvoiced tally counter exceeds an unvoiced-tally threshold;
and indicating that the unvoiced sound has ended when the unvoiced tally counter passes below the unvoiced-tally threshold.

### Claim 17 (depends on 15)

The method of claim 15 which further includes a time Tupdate, and the steps:
counting the number of times Nv that the integral signal exceeds the integral-signal threshold during Tupdate;
counting the number of times Nu that the differential signal exceeds the differential-signal threshold during Tupdate;
then, at the expiration of Tupdate, adding Nv to the voiced tally counter and adding Nu to the unvoiced tally counter;
then decrementing the voiced tally counter and the unvoiced tally counter;
then indicating that a voiced sound is present when the voiced tally counter exceeds a voiced-tally threshold;
and indicating that an unvoiced sound is present when the unvoiced tally counter exceeds an unvoiced-tally threshold.

### Claim 18 (depends on 15)

The method of claim 15 which further includes the steps:
indicating that a voiced sound has started when the voiced tally counter exceeds an upper voiced-tally threshold;
indicating that a voiced sound has ended when the voiced tally counter passes below a lower voiced-tally threshold that is lower than the upper voiced-tally threshold;
indicating that an unvoiced sound has started when the unvoiced tally counter exceeds an upper unvoiced-tally threshold;
and indicating that an unvoiced sound has ended when the unvoiced tally counter passes below a lower unvoiced-tally threshold that is lower than the upper unvoiced-tally threshold.

### Claim 19 (depends on 15)

The method of claim 15 which further includes an output signal that can be set to a first state, a second state, or a third state, said method further including the steps:
setting the output signal to the first state while the voiced tally counter exceeds a voiced-tally threshold;
setting the output signal to the second state while the voiced tally counter remains below the voiced-tally threshold and the unvoiced tally counter remains below an unvoiced-tally threshold;
and setting the output signal to the third state while the unvoiced tally counter exceeds the unvoiced-tally threshold.

### Claim 20 (depends on 4)

The method of claim 4 which further includes either:
suppressing the detection of voiced sounds while an unvoiced sound is detected;
or suppressing the detection of unvoiced sounds while a voiced sound is detected.
