SoftwareDocumentation

List of parameters

The parameters used in ManyEars are listed below. Each parameter is assigned a default value. Some parameters need to be changed, some are set to the default value most of the time, and others should not be modified. This is specified in the description of each parameter.

General

Parameters Type Default value
GLOBAL_FRAMESIZE unsigned int 1024
GLOBAL_LONGFRAMESIZE unsigned int 2048
GLOBAL_OVERLAP float 0.5
GLOBAL_C float 343.0
GLOBAL_FS unsigned int 48000
GLOBAL_BSIZE unsigned int 3
GLOBAL_WINDOW0 float 0.25
GLOBAL_WINDOW1 float 0.50
GLOBAL_WINDOW2 float 0.25
GLOBAL_SPHERE_NUMBERLEVELS float 4
GLOBAL_MICSNUMBER float 8
GLOBAL_NBPARTICLES float 500
P_GEN_DYNSOURCES unsigned int 8

Geometry

Parameters Type Default value
P_GEO_MICS_MIC1_X float -0.16
P_GEO_MICS_MIC1_Y float +0.16
P_GEO_MICS_MIC1_Z float -0.16
P_GEO_MICS_MIC1_GAIN float 1.00
P_GEO_MICS_MIC2_X float -0.16
P_GEO_MICS_MIC2_Y float +0.16
P_GEO_MICS_MIC2_Z float +0.16
P_GEO_MICS_MIC2_GAIN float 1.00
P_GEO_MICS_MIC3_X float +0.16
P_GEO_MICS_MIC3_Y float +0.16
P_GEO_MICS_MIC3_Z float -0.16
P_GEO_MICS_MIC3_GAIN float 1.00
P_GEO_MICS_MIC4_X float +0.16
P_GEO_MICS_MIC4_Y float +0.16
P_GEO_MICS_MIC4_Z float +0.16
P_GEO_MICS_MIC4_GAIN float 1.00
P_GEO_MICS_MIC5_X float -0.16
P_GEO_MICS_MIC5_Y float -0.16
P_GEO_MICS_MIC5_Z float -0.16
P_GEO_MICS_MIC5_GAIN float 1.00
P_GEO_MICS_MIC6_X float -0.16
P_GEO_MICS_MIC6_Y float -0.16
P_GEO_MICS_MIC6_Z float +0.16
P_GEO_MICS_MIC6_GAIN float 1.00
P_GEO_MICS_MIC7_X float +0.16
P_GEO_MICS_MIC7_Y float -0.16
P_GEO_MICS_MIC7_Z float -0.16
P_GEO_MICS_MIC7_GAIN float 1.00
P_GEO_MICS_MIC8_X float +0.16
P_GEO_MICS_MIC8_Y float -0.16
P_GEO_MICS_MIC8_Z float +0.16
P_GEO_MICS_MIC8_GAIN float 1.00

Preprocessing

Stationnary Noise Estimator
Parameters Type Default value
P_MCRA_ALPHAS float 0.95
P_MCRA_ALPHAP float 0.2
P_MCRA_ALPHAD float 0.95
P_MCRA_L unsigned int 150
P_MCRA_DELTA float 0.01
Signal-to-Noise Estimator
Parameters Type Default value
P_MICST_ALPHAD float 0.1
P_MICST_GAMMA float 0.3
P_MICST_DELTA float 1.0

Localization

Beamformer
Parameters Type Default value
P_BF_MAXSOURCES unsigned int 4
P_BF_FILTERRANGE unsigned int 5
P_BF_RESETRANGE unsigned int 5
P_BF_ET float 600.0

Tracking

Particle Filter
Parameters Type Default value
P_FILTER_STDDEVIATION float 0.2
P_FILTER_ALPHASTOP float 2.00
P_FILTER_BETASTOP float 0.04
P_FILTER_ALPHACONST float 0.05
P_FILTER_BETACONST float 0.50
P_FILTER_ALPHAEXC float 0.50
P_FILTER_BETASEXC float 0.20
P_FILTER_INERTIAX float 1.00
P_FILTER_INERTIAY float 1.00
P_FILTER_INERTIAZ float 1.00
P_FILTER_DELTAT float 0.008
P_FILTER_STATEUPDT float 0.03
P_FILTER_NEWSTOP float 0.50
P_FILTER_NEWCONST float 0.20
P_FILTER_NEWEXC float 0.30
P_FILTER_AJT_AJTM1 float 0.70
P_FILTER_AJT_NOTAJTM1 float 0.30
P_FILTER_P0 float 0.50
P_FILTER_RSTHRESHOLD float 0.70
P_FILTER_BUFFERSIZE unsigned int 1
Mixture
Parameters Type Default value
P_MIXTURE_PNEW float 0.005
P_MIXTURE_PFALSE float 0.05
P_MIXTURE_NEWTHRESHOLD float 0.5
P_MIXTURE_CONFIRMEXISTS float 0.7
P_MIXTURE_CONFIRMCOUNTTS float 0.5
P_MIXTURE_CONFIRMCOUNT unsigned int 50
P_MIXTURE_NEWANGLE float 1.0
P_MIXTURE_CUMULATIVETIMEPROB unsigned int 50
P_MIXTURE_TOBSPROB float 0.95
P_MIXTURE_CUMULATIVETIME1 unsigned int 25
P_MIXTURE_TOBS1 float 0.5
P_MIXTURE_CUMULATIVETIME1 unsigned int 50
P_MIXTURE_TOBS1 float 0.95

Separation

Geometric Source Separation
Parameters Type Default value
P_GSS_SOURCEDISTANCE float 3.0
P_GSS_MU float 0.001
P_GSS_LAMBDA float 0.5
Post-filter
Parameters Type Default value
P_POSTFILTER_ALPHAS float 0.3
P_POSTFILTER_ETA float 0.3
P_POSTFILTER_GAMMA float 0.3
P_POSTFILTER_DELTA float 1.0
P_POSTFILTER_TETA_LOCAL float 0.01
P_POSTFILTER_TETA_GLOBAL float 0.1
P_POSTFILTER_TETA_FRAME float 0.5
P_POSTFILTER_ALPHAZETA float 0.3
P_POSTFILTER_MAXQ float 0.9
P_POSTFILTER_GMIN float 0.9
P_POSTFILTER_LOCALWINDOWSIZE unsigned int 3
P_POSTFILTER_GLOBALWINDOWSIZE unsigned int 31
P_POSTFILTER_ALPHAPMIN float 0.07

Output

Parameters Type Default value
P_OUT_GAIN float 5.0
P_OUT_WAVEHEADERSIZE unsigned int 100
P_OUT_MINDURATION unsigned int 5 * GLOBAL_FS
P_OUT_INTERVALDURATION unsigned int 1 * GLOBAL_FS

Description of parameters

General

GLOBAL_FRAMESIZE
The size in samples of the frame used for generating the Short-Time Fourier Transform (STFT). The value needs to be a power of 2. Do not modify.
GLOBAL_LONGFRAMESIZE
Twice the number of samples of GLOBAL_FRAMESIZE. For internal use only. Do not modify.
GLOBAL_OVERLAP
The overlap between two consecutives frames. An overlap of 0.5 means on a frame of 1024 samples means a hop size of 512 samples. Do not modify.
GLOBAL_C
The speed of sound in meters/second. This constant may change slightly according to temperature, humidity and other factors. There is usually no need to change this parameter.
GLOBAL_FS
The sample rate in samples / second. Do not modify.
GLOBAL_BSIZE
The size of the hann window used for many filtering operations. For internal use only. Do not modify.
GLOBAL_WINDOW0
The first coefficient of a 3-sample hann window used for many filtering operations. For internal use only. Do not modify.
GLOBAL_WINDOW1
The second coefficient of a 3-sample hann window used for many filtering operations. For internal use only. Do not modify.
GLOBAL_WINDOW2
The third coefficient of a 3-sample hann window used for many filtering operations. For internal use only. Do not modify.
GLOBAL_SPHERE_NUMBERLEVELS
The resolution of the virtual unit sphere generated to scan for potential source positions. The number of points on the sphere is given by (20 * (4^N)) / 2 + 2, where N stands for the number of levels. When N = 0, there are 12 points, when N = 1, there are 42 points, when N = 2, there are 162 points, when N = 3, there are 642 points, when N = 4, there are 2562 points, and so on. Decreasing the number of points will decrease the sphere resolution, and will affect the precision of the localization algorithm. If the resolution is too low, some missed detection will also occur. Decreasing the resolution improve the processing speed. Increasing the resolution lead to better resolution, but also increases the computational load. There is usually no need to change this parameter.
GLOBAL_MICSNUMBER
The number of microphones in the microphone array. There must be a least two microphones. Increasing the number of microphones will increase the computational load according to a complexity order O(N^2).
GLOBAL_NBPARTICLES
The number of particles in the particle filters. If this is decreased too much, there will not be enough particles to perform efficient density estimation. Increasing this number will improve the approximation but also increase computing power. There is usually no need to change this parameter.
P_GEN_DYNSOURCES
The maximum number of sources that can be tracked and separated simultaneously.

Geometry

P_GEO_MICS_MIC1_X
The x-coordinate of microphone 1 in meter. The origin of the xyz axes is set by the user but must be the same for all microphone coordinates. The center of mass of the microphone array is usually a good choice of origin.
P_GEO_MICS_MIC1_Y
The y-coordinate of microphone 1 in meter. The origin of the xyz axes is set by the user but must be the same for all microphone coordinates. The center of mass of the microphone array is usually a good choice of origin.
P_GEO_MICS_MIC1_Z
The z-coordinate of microphone 1 in meter. The origin of the xyz axes is set by the user but must be the same for all microphone coordinates. The center of mass of the microphone array is usually a good choice of origin.
P_GEO_MICS_MIC1_GAIN
A gain applied directly to the signal that comes out of microphone 1. Usually there is no need to modify this gain unless a microphone needs to be turned off, in which case the gain is set to 0.
P_GEO_MICS_MIC2_X
The x-coordinate of microphone 2 in meter. The origin of the xyz axes is set by the user but must be the same for all microphone coordinates. The center of mass of the microphone array is usually a good choice of origin.
P_GEO_MICS_MIC2_Y
The y-coordinate of microphone 2 in meter. The origin of the xyz axes is set by the user but must be the same for all microphone coordinates. The center of mass of the microphone array is usually a good choice of origin.
P_GEO_MICS_MIC2_Z
The z-coordinate of microphone 2 in meter. The origin of the xyz axes is set by the user but must be the same for all microphone coordinates. The center of mass of the microphone array is usually a good choice of origin.
P_GEO_MICS_MIC2_GAIN
A gain applied directly to the signal that comes out of microphone 2. Usually there is no need to modify this gain unless a microphone needs to be turned off, in which case the gain is set to 0.
P_GEO_MICS_MIC3_X
The x-coordinate of microphone 3 in meter. The origin of the xyz axes is set by the user but must be the same for all microphone coordinates. The center of mass of the microphone array is usually a good choice of origin.
P_GEO_MICS_MIC3_Y
The y-coordinate of microphone 3 in meter. The origin of the xyz axes is set by the user but must be the same for all microphone coordinates. The center of mass of the microphone array is usually a good choice of origin.
P_GEO_MICS_MIC3_Z
The z-coordinate of microphone 3 in meter. The origin of the xyz axes is set by the user but must be the same for all microphone coordinates. The center of mass of the microphone array is usually a good choice of origin.
P_GEO_MICS_MIC3_GAIN
A gain applied directly to the signal that comes out of microphone 3. Usually there is no need to modify this gain unless a microphone needs to be turned off, in which case the gain is set to 0.
P_GEO_MICS_MIC4_X
The x-coordinate of microphone 4 in meter. The origin of the xyz axes is set by the user but must be the same for all microphone coordinates. The center of mass of the microphone array is usually a good choice of origin.
P_GEO_MICS_MIC4_Y
The y-coordinate of microphone 4 in meter. The origin of the xyz axes is set by the user but must be the same for all microphone coordinates. The center of mass of the microphone array is usually a good choice of origin.
P_GEO_MICS_MIC4_Z
The z-coordinate of microphone 4 in meter. The origin of the xyz axes is set by the user but must be the same for all microphone coordinates. The center of mass of the microphone array is usually a good choice of origin.
P_GEO_MICS_MIC4_GAIN
A gain applied directly to the signal that comes out of microphone 4. Usually there is no need to modify this gain unless a microphone needs to be turned off, in which case the gain is set to 0.
P_GEO_MICS_MIC5_X
The x-coordinate of microphone 5 in meter. The origin of the xyz axes is set by the user but must be the same for all microphone coordinates. The center of mass of the microphone array is usually a good choice of origin.
P_GEO_MICS_MIC5_Y
The y-coordinate of microphone 5 in meter. The origin of the xyz axes is set by the user but must be the same for all microphone coordinates. The center of mass of the microphone array is usually a good choice of origin.
P_GEO_MICS_MIC5_Z
The z-coordinate of microphone 5 in meter. The origin of the xyz axes is set by the user but must be the same for all microphone coordinates. The center of mass of the microphone array is usually a good choice of origin.
P_GEO_MICS_MIC5_GAIN
A gain applied directly to the signal that comes out of microphone 5. Usually there is no need to modify this gain unless a microphone needs to be turned off, in which case the gain is set to 0.
P_GEO_MICS_MIC6_X
The x-coordinate of microphone 6 in meter. The origin of the xyz axes is set by the user but must be the same for all microphone coordinates. The center of mass of the microphone array is usually a good choice of origin.
P_GEO_MICS_MIC6_Y
The y-coordinate of microphone 6 in meter. The origin of the xyz axes is set by the user but must be the same for all microphone coordinates. The center of mass of the microphone array is usually a good choice of origin.
P_GEO_MICS_MIC6_Z
The z-coordinate of microphone 6 in meter. The origin of the xyz axes is set by the user but must be the same for all microphone coordinates. The center of mass of the microphone array is usually a good choice of origin.
P_GEO_MICS_MIC6_GAIN
A gain applied directly to the signal that comes out of microphone 6. Usually there is no need to modify this gain unless a microphone needs to be turned off, in which case the gain is set to 0.
P_GEO_MICS_MIC7_X
The x-coordinate of microphone 7 in meter. The origin of the xyz axes is set by the user but must be the same for all microphone coordinates. The center of mass of the microphone array is usually a good choice of origin.
P_GEO_MICS_MIC7_Y
The y-coordinate of microphone 7 in meter. The origin of the xyz axes is set by the user but must be the same for all microphone coordinates. The center of mass of the microphone array is usually a good choice of origin.
P_GEO_MICS_MIC7_Z
The z-coordinate of microphone 7 in meter. The origin of the xyz axes is set by the user but must be the same for all microphone coordinates. The center of mass of the microphone array is usually a good choice of origin.
P_GEO_MICS_MIC7_GAIN
A gain applied directly to the signal that comes out of microphone 7. Usually there is no need to modify this gain unless a microphone needs to be turned off, in which case the gain is set to 0.
P_GEO_MICS_MIC8_X
The x-coordinate of microphone 8 in meter. The origin of the xyz axes is set by the user but must be the same for all microphone coordinates. The center of mass of the microphone array is usually a good choice of origin.
P_GEO_MICS_MIC8_Y
The y-coordinate of microphone 8 in meter. The origin of the xyz axes is set by the user but must be the same for all microphone coordinates. The center of mass of the microphone array is usually a good choice of origin.
P_GEO_MICS_MIC8_Z
The z-coordinate of microphone 8 in meter. The origin of the xyz axes is set by the user but must be the same for all microphone coordinates. The center of mass of the microphone array is usually a good choice of origin.
P_GEO_MICS_MIC8_GAIN
A gain applied directly to the signal that comes out of microphone 8. Usually there is no need to modify this gain unless a microphone needs to be turned off, in which case the gain is set to 0.

Preprocessing

Stationnary Noise Estimator
P_MCRA_ALPHAS
This parameter is used to smooth spectrum in time according to the Minima Controlled Recursive Averaging (MCRA) technique. There is usually no need to change this parameter.
P_MCRA_ALPHAP
This parameter is used to smooth the conditional signal presence probability estimator in time according to the Minima Controlled Recursive Averaging (MCRA) technique. There is usually no need to change this parameter.
P_MCRA_ALPHAD
This parameter is used to smooth the variance of noise estimator in time according to the Minima Controlled Recursive Averaging (MCRA) technique. There is usually no need to change this parameter.
P_MCRA_L
This is the number of frames used to compute a local minima for each frequency bin. There is usually no need to change this parameter.
P_MCRA_DELTA
This is the threshold used to classify the ratio between local energy of noisy speech and its local minimum as speech or noise. There is usually no need to change this parameter.
Signal-to-Noise Estimator
P_MICST_ALPHAD
This parameter is used to smooth the signal-to-noise ratio estimator in time according. There is usually no need to change this parameter.
P_MICST_GAMMA
This parameter stands for the reverberation time of the room. There is usually no need to change this parameter.
P_MICST_DELTA
This parameter stands for the signal-to-reverberatan ratio. There is usually no need to change this parameter.

Localization Parameters

Beamformer Parameters
P_BF_MAXSOURCES
The number of potential sources that are generated from the localization module. This number does not correspond to the maximum number of sources to be tracked, but rather to the number of scans performed around the sphere to find potential sources. Since the sources are sporadically active over time, it is possible to track more sources than the number of potential sources. Increasing the number of scan usually provide the tracking module with more confident values about the active sources, but also significantly increase the computational load. The value can also be reduced to one for a single scan. This will not affect performances when a single source is tracked and will slightly reduce performances when many sources are tracked. Reducing this value to 1 will improve significantly real-time performances of the system.
P_BF_FILTERRANGE
This parameter deals with the fact that the microphones are never perfectly calibrated in terms of positions. The theroetical time-delay between a pair of microphone for a given signal may be slightly different than the measured time-delay. Given this parameter is represented by the variable d and the exact delay is given by the variable d0, then the system will look for best delay match over the range (d0-d) to (d0+d). There is usually no need to change this parameter.
P_BF_RESETRANGE
When there is more than one scan over the sphere, the delays associated to the previously found potential sources are reset to zero. The delays close to the exact delays are also reset. Given this parameter is represented by the variable r and the exact delay is given by the variable d0, then the system will reset the delays over the range (d0-r) to (d0+r). There is usually no need to change this parameter.
P_BF_ET
This parameter stands for the energy threshold to generate the probability associated to each potential source. Decreasing this level may improve source detection and tracking, but may also introduce false detections and bad tracking. Increasing this level may improve robustness to noise, but may also introduce missed detection and bad tracking.

Tracking

Particle Filter
P_FILTER_STDDEVIATION
When a new source is added, the particles sample a normal distribution, which means correspond to the position of the new source, and the standard deviation to this parameter. There is usually no need to change this parameter.
P_FILTER_ALPHASTOP
This parameters stands for the alpha constant in the damping model used to represent a stationary source. There is usually no need to change this parameter.
P_FILTER_BETASTOP
This parameters stands for the beta constant in the damping model used to represent a stationary source. There is usually no need to change this parameter.
P_FILTER_ALPHACONST
This parameters stands for the alpha constant in the damping model used to represent a source moving with a constant velocity. There is usually no need to change this parameter.
P_FILTER_BETACONST
This parameters stands for the beta constant in the damping model used to represent a source moving with a constant velocity. There is usually no need to change this parameter.
P_FILTER_ALPHANEW
This parameters stands for the alpha constant in the damping model used to represent an accelerated source. There is usually no need to change this parameter.
P_FILTER_BETANEW
This parameters stands for the beta constant in the damping model used to represent an accelerated source. There is usually no need to change this parameter.
P_FILTER_INERTIAX
This parameter is used to speed up or slow down the motion of a source in the x-plane. When a source motion is limited on a given plane (e.g. the floor), this can improve the tracking performance. However, usually this parameter is set to one and does not influence the tracking. There is usually no need to change this parameter.
P_FILTER_INERTIAY
This parameter is used to speed up or slow down the motion of a source in the y-plane. When a source motion is limited on a given plane (e.g. the floor), this can improve the tracking performance. However, usually this parameter is set to one and does not influence the tracking. There is usually no need to change this parameter.
P_FILTER_INERTIAZ
This parameter is used to speed up or slow down the motion of a source in the z-plane. When a source motion is limited on a given plane (e.g. the floor), this can improve the tracking performance. However, usually this parameter is set to one and does not influence the tracking. There is usually no need to change this parameter.
P_FILTER_DELTAT
This parameter represents the interval of time between each update of the damping model. There is usually no need to change this parameter.
P_FILTER_STATEUPDT
Once in a while particles must change from a state (stopped, constant velocity particle and acceleration) to another state. This parameters stands for the percentage of particles which state is modified at each iteration. There is usually no need to change this parameter.
P_FILTER_NEWSTOP
This is the percentage of particles with a stopped state. It is important to ensure the sum of the parameters P_FILTER_NEWSTOP, P_FILTER_NEWCONST and P_FILTER_NEWEXC is 1.0. If the application suggests that the sound sources are static in space, this parameter may be increased (and the others decrease) to improve performances.
P_FILTER_NEWCONST
This is the percentage of particles with a constant velocity state. It is important to ensure the sum of the parameters P_FILTER_NEWSTOP, P_FILTER_NEWCONST and P_FILTER_NEWEXC is 1.0. If the application suggests that the sound sources are moving a lot in space, this parameter may be increased (and the P_FILTER_NEWSTOP decrease while P_FILTER_NEWEXC can be kept constant) to improve performances.
P_FILTER_NEWEXC
This is the percentage of particles with an acceleration state. It is important to ensure the sum of the parameters P_FILTER_NEWSTOP, P_FILTER_NEWCONST and P_FILTER_NEWEXC is 1.0.
P_FILTER_AJT_AJTM1
This parameter stands for the transition probability in a Hidden Markov Model that the source was active and is still active. Do not modify.
P_FILTER_AJT_NOTAJTM1
This parameter stands for the transition probability in a Hidden Markov Model that the source was not active and becomes active. Do not modify.
P_FILTER_P0
This parameter stands for the probability a source is not observed even if it exists. Do not modify.
P_FILTER_RSTHRESHOLD
This parameter is the resampling threshold for each particle filter. Resampling is required to prevent the filter to degenerate to a single non-zero particle. Do not modify.
P_FILTER_BUFFERSIZE
This parameter is the size of the buffer in samples to add delay during the parameter estimation step to produce smooth trajectories. Do not modify.
Mixture
P_MIXTURE_PNEW
This parameter stands for the probability a new source appears. Do not modify.
P_MIXTURE_PFALSE
This parameter stands for the probability a false detection occurs. Do not modify.
P_MIXTURE_NEWTHRESHOLD
This parameter is the threshold to reach in order to track a new source. By decreasing this value, the system becomes more sensitive to new source, but also more vulnerable to false detections.
P_MIXTURE_CONFIRMEXISTS
This parameter stands for the threshold to reach during a single iteration once a new source is detected to confirm this source really exists and start tracking this source. There is usually no need to change this parameter.
P_MIXTURE_CONFIRMCOUNTTS
This parameter stands for the threshold to reach during a many iterations (the number of iterations is set by P_MIXTURE_CONFIRMCOUNT) once a new source is detected to confirm this source really exists and start tracking this source. There is usually no need to change this parameter.
P_MIXTURE_CONFIRMCOUNT
This parameter stands for the number of iterations for the probability to be above the threshold P_MIXTURE_CONFIRMCOUNTTS once a new source is detected to confirm this source really exists and start tracking this source. There is usually no need to change this parameter.
P_MIXTURE_NEWANGLE
This parameter is no longer used in the new versions of the system. Do not modify.
P_MIXTURE_CUMULATIVETIMEPROB
Once a new source is being tracked, there is an additional probation time before displaying the source as being tracked. If the source is not observed (i.e. is below the threshold P_MIXTURE_TOBSPROB) during P_MIXTURE_CUMULATIVETIMEPROB iterations, then it is deleted. There is usually no need to change this parameter.
P_MIXTURE_TOBSPROB
Once a new source is being tracked, there is an additional probation time before displaying the source as being tracked. If the source is not observed (i.e. is below the threshold P_MIXTURE_TOBSPROB) during P_MIXTURE_CUMULATIVETIMEPROB iterations, then it is deleted. There is usually no need to change this parameter.
P_MIXTURE_CUMULATIVETIME1
Once a new source is being tracked, it can be deleted if it is not observed (i.e. is below the threshold P_MIXTURE_TOBS1) during P_MIXTURE_CUMULATIVETIME1 iterations. Decreasing the threshold makes it harder to delete sources once they stop being active. Increasing the threshold may make the source being deleted even if it is still active.
P_MIXTURE_TOBS1
Once a new source is being tracked, it can be deleted if it is not observed (i.e. is below the threshold P_MIXTURE_TOBS1) during P_MIXTURE_CUMULATIVETIME1 iterations. Decreasing the threshold makes it harder to delete sources once they stop being active. Increasing the threshold may make the source being deleted even if it is still active.
P_MIXTURE_CUMULATIVETIME2
Once a new source is being tracked, it can be deleted if it is not observed (i.e. is below the threshold P_MIXTURE_TOBS2) during P_MIXTURE_CUMULATIVETIME2 iterations. Decreasing the threshold makes it harder to delete sources once they stop being active. Increasing the threshold may make the source being deleted even if it is still active.
P_MIXTURE_TOBS2
Once a new source is being tracked, it can be deleted if it is not observed (i.e. is below the threshold P_MIXTURE_TOBS2) during P_MIXTURE_CUMULATIVETIME2 iterations. Decreasing the threshold makes it harder to delete sources once they stop being active. Increasing the threshold may make the source being deleted even if it is still active.

Separation

Geometric Source Separation
P_GSS_SOURCEDISTANCE
This parameter stands for the approximate source distance of the sources around the microphone array. The distance is not critical but an approximation of the usual distance between the sound source and the microphone array may improve the delay-and-sum beamformer separation. There is usually no need to change this parameter.
P_GSS_MU
This parameter stands for the update rate for the sound source separation algorithm. This rate must be chosen carefully: if the value is too large, this will lead to divergence of the separation algorithm. If the rate is set to zero, there is no separation based on the decorrelation of the signals.
P_GSS_LAMBDA
This parameter stands for the regularisation term to make the demixing matrix as small as possible. There is usually no need to change this parameter.
Post-filter
P_POSTFILTER_ALPHAS
This parameter is used to smooth spectrum in time according to the Minima Controlled Recursive Averaging (MCRA) technique. There is usually no need to change this parameter.
P_POSTFILTER_ETA
This parameter stands for the factor of reduction of the interferences after separation. There is usually no need to change this parameter.
P_POSTFILTER_GAMMA
This parameter stands for the reverberation time of the room. There is usually no need to change this parameter.
P_POSTFILTER_DELTA
This parameter stands for the signal-to-reverberatan ratio. There is usually no need to change this parameter.
P_POSTFILTER_TETA_LOCAL
This parameter stands for a soft decision threshold used to detect speech activity in a local window. Do not modify.
P_POSTFILTER_TETA_GLOBAL
This parameter stands for a soft decision threshold used to detect speech activity in a global window. Do not modify.
P_POSTFILTER_TETA_FRAME
This parameter stands for a soft decision threshold used to detect speech activity in a frame. Do not modify.
P_POSTFILTER_ALPHAZETA
This parameter stands for an adaptation rate for the recursive estimation of the a priori SNR. Do not modify.
P_POSTFILTER_MAXQ
This parameter stands for the maximum a priori probability of speech absence. Do not modify.
P_POSTFILTER_GMIN
This parameter stands for the minimum gain when speech is absent. Do not modify.
P_POSTFILTER_LOCALWINDOWSIZE
This parameter stands for the number of bins in a local window. Do not modify.
P_POSTFILTER_GLOBALWINDOWSIZE
This parameter stands for the number of bins in a global window. Do not modify.
P_POSTFILTER_ALPHAPMIN
This parameter stands for the minimum value of the update rate of the a priori SNR. Do not modify.

Output

P_OUT_GAIN
This parameter stands for the gain of the separated signals. A large gain may lead to clipping.
P_OUT_WAVEHEADERSIZE
This parameter stands for the number of bytes in the header of the wave files. Do not modify.
P_OUT_MINDURATION
This parameter is used for future developments and can be ignored. Do not modify.
P_OUT_INTERVALDURATION
This parameter is used for future developments and can be ignored. Do not modify.

Related

ManyEars-Wiki: Main_Page

Get latest updates about Open Source Projects, Conferences and News.

Sign up for the SourceForge newsletter:





No, thanks