我的Bilibili频道：香芋派Taro
我的个人博客：taropie0224.github.io
我的公众号：香芋派的烘焙坊
我的音频技术交流群：1136403177
我的个人微信：JazzyTaroPie

Stereo Image Widening

Github：https://github.com/TaroPie0224/TaroStereoWidth

立体声扩展基于mid-side processing，如果为了让信号听起来更宽，side通道会相对于mid通道增强，如果为了让信号听起来更窄，mid通道会相对于side通道增强。

1
2
3
4
5
6
7
8
9
10
11
12
13
14
// 首先我们需要获得左右两个声道，分别为left和right
// Encoding，通过左右声道得到sides和mid
sides = (1/2) * (left - right);
mid = (1/2) * (left + right);

// Processing，width的范围是0～2
width = 1;
sides = width * sides;
mid = (2 - width) * mid;

// Decoding
newLeft = mid + sides;
newRight = mid - sides;

下面给出JUCE中processBlock中的dsp部分代码

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
void TaroStereoWidthAudioProcessor::processBlock (juce::AudioBuffer<float>& buffer, juce::MidiBuffer& midiMessages)
{
    juce::ScopedNoDenormals noDenormals;
    auto totalNumInputChannels  = getTotalNumInputChannels();
    auto totalNumOutputChannels = getTotalNumOutputChannels();
    
    auto currentGain = apvts.getRawParameterValue("Gain")->load();
    auto currentWidth = apvts.getRawParameterValue("Width")->load();
    
    for (auto i = totalNumInputChannels; i < totalNumOutputChannels; ++i)
        buffer.clear (i, 0, buffer.getNumSamples());

    juce::dsp::AudioBlock<float> block {buffer};
    
    for (int sample = 0; sample < block.getNumSamples(); ++sample)
    {
        // Encoding，通过左右声道得到sides和mid
        auto* channelDataLeft = buffer.getWritePointer(0);
        auto* channelDataRight = buffer.getWritePointer(1);
        
        auto sides = 0.5 * (channelDataLeft[sample] - channelDataRight[sample]);
        auto mid = 0.5 * (channelDataLeft[sample] + channelDataRight[sample]);
        
        // Processing，width的范围是0～2
        sides = currentWidth * sides;
        mid = (2 - currentWidth) * mid;
        
        // Decoding
        channelDataLeft[sample] = (mid + sides) * currentGain;
        channelDataRight[sample] = (mid - sides) * currentGain;
    }
}