Commit 7074a592 authored by PLAINSI's avatar PLAINSI
Browse files

Fix lfilter input dimensions

parent 107084b7

item_gen_configs/FOA_CONFIG.yml

+21 −6
Original line number Diff line number Diff line
@@ -56,9 +56,24 @@ scenes: 
        IR: ["AmbiX_036x033y_0.01m.wav", "AmbiX_-70x035y_0.01m.wav"]

        overlap: 1.0

        

    #a2: 

    #    name: "G6S2.wav"

    #    description: "Two speakers sitting in a car."

    #    source: ["fa1.wav", "ma1.wav"]

    #    IR: ["AmbiX_036x033y_0.01m.wav", "AmbiX_-70x035y_0.01m.wav"]

    #    overlap: 1.0
 
    a2: 

        name: "G6S2.wav"

        description: "Two speakers sitting in a car."

        source: ["fa1.wav", "ma1.wav"]

        IR: ["AmbiX_-70x035y_0.01m.wav", "AmbiX_036x033y_0.01m.wav"]

        overlap: 1.0



    a3: 

        name: "G2S1.wav"

        description: "Two speakers sitting in a car."

        source: ["fa1.wav", "ma1.wav"]

        IR: ["AmbiX_000x005y_0.1m.wav", "AmbiX_022x-38y_0.01m.wav"]

        overlap: -1.0

        

    a4: 

        name: "G2S2.wav"

        description: "Two speakers sitting in a car."

        source: ["fa1.wav", "ma1.wav"]

        IR: ["AmbiX_022x-38y_0.01m.wav", "AmbiX_000x005y_0.1m.wav"]

        overlap: -1.0

ivas_processing_scripts/generation/process_foa_items.py

+7 −6
Original line number Diff line number Diff line
@@ -88,7 +88,9 @@ def filter_one( 


    tmp_IR.audio = tmp_IR.audio * align

    output = copy(input)

    output.audio[0] = ssg.lfilter(tmp_IR.audio[0],1,input.audio[0])

    intranspose = input.audio.transpose()

    outfilt = ssg.lfilter(tmp_IR.audio,[1.0],intranspose)

    output.audio = outfilt.transpose()



    return output
@@ -120,22 +122,22 @@ def filter_foa( 
    IR_w = copy(foa_IR)

    IR_w.name = "MONO"

    IR_w.num_channels = 1

    IR_w.audio = np.reshape(foa_IR.audio[:, 0], (-1, 1))

    IR_w.audio = foa_IR.audio[:, 0] # np.reshape(foa_IR.audio[:, 0], (-1, 1))



    IR_x = copy(foa_IR)

    IR_x.name = "MONO"

    IR_x.num_channels = 1

    IR_x.audio = np.reshape(foa_IR.audio[:, 1], (-1, 1))

    IR_x.audio = foa_IR.audio[:, 1] # np.reshape(foa_IR.audio[:, 1], (-1, 1))



    IR_y = copy(foa_IR)

    IR_y.name = "MONO"

    IR_y.num_channels = 1

    IR_y.audio = np.reshape(foa_IR.audio[:, 2], (-1, 1))

    IR_y.audio = foa_IR.audio[:, 2] # np.reshape(foa_IR.audio[:, 2], (-1, 1))



    IR_z = copy(foa_IR)

    IR_z.name = "MONO"

    IR_z.num_channels = 1

    IR_z.audio = np.reshape(foa_IR.audio[:, 3], (-1, 1))

    IR_z.audio = foa_IR.audio[:, 3] # np.reshape(foa_IR.audio[:, 3], (-1, 1))



    # calculate the scaling (multiplicative) factor such that the maximum gain of the IR filter across all frequencies is 0dB

    if align is None:
@@ -215,7 +217,6 @@ def generate_foa_items( 
        for i in range(N_sources):

            # parse parameters from the scene description

            source_file = np.atleast_1d(scene["source"])[i]

            print("source file: {}".format(source_file))

            IR_file = np.atleast_1d(scene["IR"])[i]



            logger.info(f"Convolving {source_file} with {IR_file}")