WavLMSEROptions

Options & Configuration · AiDotNet.Audio.Emotion

Configuration options for the WavLM-SER speech emotion recognition model.

Inheritance: ModelOptions → WavLMSEROptions

For Beginners

WavLM-SER takes a model that first learned to understand speech in general, then teaches it to recognize emotions specifically. Because WavLM already understands the deep patterns of human speech, it picks up on subtle emotional cues that simpler models miss—like the slight tremor in a fearful voice or the rise in pitch when someone is excited.

How It Works

WavLM-SER fine-tunes the WavLM self-supervised model (Chen et al., 2022) for speech emotion recognition. WavLM’s pre-training on masked speech prediction and denoising produces robust features that, when fine-tuned for SER, achieve state-of-the-art results on IEMOCAP (weighted accuracy 71%+) and are robust to noise and recording conditions.

Constructors

Constructor Summary
WavLMSEROptions() Initializes WavLM-SER with the paper-scale base defaults.
WavLMSEROptions(WavLMSEROptions other) Creates an independent copy of an existing option set.

Properties

Property Summary
OnnxModelOptions OnnxOptions { get; set; } Gets or sets the ONNX runtime options.
bool IncludeArousalValence { get; set; } Gets or sets whether to include arousal/valence estimation.
double DropoutRate { get; set; } Gets or sets the dropout rate.
double LearningRate { get; set; } Gets or sets the downstream emotion-recognition learning rate.
double WeightDecay { get; set; } Gets or sets the decoupled AdamW weight decay.
int FeatureEncoderDim { get; set; } Gets or sets the CNN feature encoder output dimension.
int FeatureLayerIndex { get; set; } Gets or sets which WavLM layer to use for features (-1 = weighted sum).
int FeedForwardDim { get; set; } Gets or sets the feed-forward hidden dimension.
int FftSize { get; set; } Gets or sets the FFT window size.
int HiddenDim { get; set; } Gets or sets the Transformer hidden dimension.
int HopLength { get; set; } Gets or sets the hop length between frames.
int NumAttentionHeads { get; set; } Gets or sets the number of attention heads.
int NumClasses { get; set; } Gets or sets the number of emotion classes.
int NumLayers { get; set; } Gets or sets the number of Transformer layers.
int NumMels { get; set; } Gets or sets the number of mel filterbank channels.
int SampleRate { get; set; } Gets or sets the expected audio sample rate in Hz.
string ModelPath { get; set; } Gets or sets the path to the ONNX model file.
string Variant { get; set; } Gets or sets the model variant (“base” or “large”).
string[] EmotionLabels { get; set; } Gets or sets the emotion label names.