# US 2019/0378531 A1 · Audio processing device and a method for estimating a signal-to-noise-ratio of a sound signal

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

- Number: US 2019/0378531 A1
- Document ID: US20190378531A1
- Status: Published application
- Kind: A1
- Application number: 16/546,928
- Filed: 2019-08-21
- Published: 2019-12-12
- Earliest claimed: 2016-05-30
- Assignees: Oticon A/S
- Applicants: Oticon A/S
- Inventors: Jesper JENSEN; Andreas Thelander BERTELSEN; Michael Syskind PEDERSEN
- Web page: https://patentgrep.com/patent/US20190378531A1

## CPC

- G10L 21/0208: Noise filtering
- G10L 21/0232: Processing in the frequency domain
- G10L 15/20: Speech recognition techniques specially adapted for robustness in adverse environments, e.g. in noise, of stress induced speech (G10L21/02 takes precedence)
- G10L 19/26: Pre-filtering or post-filtering
- G10L 25/30: using neural networks
- 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

An audio processing device comprises a) at least one input unit for providing a time-frequency representation Y(k,n) of an electric input signal representing sound consisting of target speech and noise signal components, where k and n are frequency band and time frame indices, respectively, b) a noise reduction system configured to b1) determine a first signal to noise ratio estimate γ(k,n) of said electric input signal, and to b2) determine a second signal to noise signal ratio estimate ζ(k,n) of said electric input signal from said first signal to noise ratio estimate γ(k,n) based on a recursive algorithm providing non-linear smoothing, and wherein a determination of said one or more bias and/or smoothing parameters comprises the use of supervised learning, e.g. one or more neural networks. The invention may be used in audio processing devices, such as hearing aids, headsets, ear phones, active ear protection systems, handsfree telephone systems, mobile telephones, etc.

## Claims

### Claim 1

An audio processing device, comprising
at least one input unit for providing a time-frequency representation Y(k,n) of an electric input signal representing a time variant sound signal consisting of target speech signal components S(k,n) from a target sound source TS and noise signal components N(k,n) from other sources than the target sound source, where k and n are frequency band and time frame indices, respectively,
a noise reduction system configured to
determine a first signal to noise ratio estimate γ(k,n) of said electric input signal,
determine second signal to noise ratio estimate ζ(k,n) of said electric input signal from said first signal to noise ratio estimate γ(k,n) based on a recursive algorithm comprising a recursive loop, and to
determine said second signal to noise ratio estimate γ(k,n) by non-linear smoothing of said first signal to noise ratio estimate γ(k,n), or a parameter derived therefrom, and wherein said non-linear smoothing is controlled by one or more bias and/or smoothing parameters; and
wherein a determination of said one or more bias and/or smoothing parameters comprises the use of supervised learning implementing one or more trained neural networks.

### Claim 2 (depends on 1)

An audio processing device according to claim 1 wherein a selector is located in the recursive loop, wherein said selector is configured to select an input to determine said one or more bias and/or smoothing parameters based on a select control parameter.

### Claim 3 (depends on 2)

An audio processing device according to claim 2 wherein said select control parameter is determined using one or more neural networks.

### Claim 4 (depends on 2)

An audio processing device according to claim 2 wherein said select control parameter for a given frequency index k is determined in dependence of the first and/or the second signal to noise ratio estimates corresponding to a multitude of frequency indices.

### Claim 5 (depends on 4)

An audio processing device according to claim 4 wherein said multitude of frequency indices include one or more neighboring frequency indices.

### Claim 6 (depends on 1)

An audio processing device according to claim 1 wherein said select control parameter for a given frequency index k is additionally determined in dependence of inputs from one or more detectors.

### Claim 7 (depends on 6)

An audio processing device according to claim 6 wherein said one or more detectors comprise a general onset detector for detecting sudden changes in the time variant input sound, a wind noise detector, a voice detector, a head movement detector, a wireless transmission detector, voice detectors from microphones in other audio devices, and combinations thereof.

### Claim 8 (depends on 6)

An audio processing device according to claim 6 wherein at least one of said one or more detectors is based on binaural detection.

### Claim 9 (depends on 1)

An audio processing device according to claim 1 wherein said one or more bias and/or smoothing parameters at a given time instant n is determined in dependence of
a first maximum likelihood estimate of the second signal to noise ratio estimate at that time instant n, and/or of
an estimate of the second signal to noise ratio estimate at the previous time instant n−1, selected by said select control parameter.

### Claim 10 (depends on 1)

An audio processing device according to claim 1 configured to provide a noise reduction gain GNR in dependence of said second signal to noise ratio estimate ζ(k,n), and to apply said noise reduction gain GNR to said electric input signal or a signal derived therefrom.

### Claim 11 (depends on 1)

An audio processing device according to claim 1 comprising a filter bank comprising an analysis filter bank for providing said time-frequency representation Y(k,n) of said electric input signal.

### Claim 12 (depends on 11)

An audio processing device according to claim 11 configured to provide that said analysis filter bank is oversampled.

### Claim 13 (depends on 1)

An audio processing device according to claim 1 comprising a hearing device, e.g. a hearing aid.

### Claim 14

A method of estimating an a priori signal to noise ratio ζ(k,n) of a time-frequency representation Y(k,n) of an electric input signal representing a time variant sound signal consisting of target speech components and noise components, where k and n are frequency band and time frame indices, respectively, the method comprising
determining a first signal to noise ratio estimate γ(k,n) of said electric input signal Y(k,n);
determining a second signal to noise signal ratio estimate ζ(k,n) of said electric input signal from said first signal to noise ratio estimate γ(k,n) based on a recursive algorithm;
determining said second signal to noise ratio estimate ζ(k,n) by non-linear smoothing of said first signal to noise ratio estimate γ(k,n), or a parameter derived therefrom, and
wherein said non-linear smoothing is controlled by one or more bias and/or smoothing parameters; and
using supervised learning to determine said one or more bias and/or smoothing parameters.

### Claim 15 (depends on 14)

A method according to claim 14 further comprising selecting an input to determine said one or more bias and/or smoothing parameters based on a select control parameter;

### Claim 16 (depends on 15)

A method according to claim 15 wherein said select control parameter is determined using one or more neural networks.

### Claim 17 (depends on 15)

A method according to claim 15 wherein said select control parameter for a given frequency index k is determined in dependence of the first and/or the second signal to noise ratio estimates corresponding to a multitude of frequency indices.

### Claim 18 (depends on 14)

A method according to claim 14 comprising
providing a noise reduction gain GNR in dependence of said second signal to noise ratio estimate ζ(k,n); and
applying said noise reduction gain GNR to said electric input signal or a signal derived therefrom.

### Claim 19 (depends on 14)

A data processing system comprising a processor and program code means for causing the processor to perform the method of claim 14.

### Claim 20 (depends on 14)

A non-transitory computer readable medium storing a computer program comprising instructions which, when the program is executed by a computer, cause the computer to carry out the method of claim 14.