Add delta vector support

precise/network_runner.py

@@ -22,7 +22,7 @@ import numpy as np
 from precise.model import load_precise_model
 from precise.params import inject_params
 from precise.util import buffer_to_audio
-from precise.vectorization import vectorize_raw
+from precise.vectorization import vectorize_raw, add_deltas
 
 
 class Runner(metaclass=ABCMeta):
@@ -129,4 +129,7 @@ class Listener:
                 new_features = new_features[-len(self.mfccs):]
             self.mfccs = np.concatenate((self.mfccs[len(new_features):], new_features))
 
-        return self.runner.run(self.mfccs)
+        mfccs = self.mfccs
+        if self.pr.use_delta:
+            mfccs = add_deltas(self.mfccs)
+        return self.runner.run(mfccs)

precise/params.py

@@ -17,7 +17,7 @@ import json
 from math import floor
 import attr
 
-@attr.s
+@attr.s(frozen=True)
 class ListenerParams:
     window_t = attr.ib()  # type: float
     hop_t = attr.ib()  # type: float
@@ -27,6 +27,7 @@ class ListenerParams:
     n_mfcc = attr.ib()  # type: int
     n_filt = attr.ib()  # type: int
     n_fft = attr.ib()  # type: int
+    use_delta = attr.ib()  # type: bool
 
     @property
     def buffer_samples(self):
@@ -51,13 +52,13 @@ class ListenerParams:
 
     @property
     def feature_size(self):
-        return self.n_mfcc
+        return self.n_mfcc + self.use_delta * self.n_mfcc
 
 
 # Global listener parameters
 pr = ListenerParams(
     window_t=0.1, hop_t=0.05, buffer_t=1.5, sample_rate=16000,
-    sample_depth=2, n_mfcc=13, n_filt=20, n_fft=512
+    sample_depth=2, n_mfcc=13, n_filt=20, n_fft=512, use_delta=False
 )
 
 

precise/train_data.py

@@ -188,7 +188,7 @@ class TrainData:
         ])
 
     @staticmethod
-    def __load_files(kw_files: list, nkw_files: list, vectorizer: Callable = vectorize) -> tuple:
+    def __load_files(kw_files: list, nkw_files: list, vectorizer: Callable = None) -> tuple:
         inputs = []
         outputs = []
 

precise/vectorization.py

@@ -11,7 +11,9 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+import hashlib
 from typing import *
+import json
 
 import numpy as np
 
@@ -31,6 +33,14 @@ def vectorize_raw(audio: np.ndarray) -> np.ndarray:
     return mfcc(audio, pr.sample_rate, pr.window_t, pr.hop_t, pr.n_mfcc, pr.n_filt, pr.n_fft)
 
 
+def add_deltas(features: np.ndarray) -> np.ndarray:
+    deltas = np.zeros_like(features)
+    for i in range(1, len(features)):
+        deltas[i] = features[i] - features[i - 1]
+
+    return np.concatenate([features, deltas], -1)
+
+
 def vectorize(audio: np.ndarray) -> np.ndarray:
     """
     Args:
@@ -53,6 +63,10 @@ def vectorize(audio: np.ndarray) -> np.ndarray:
     return features
 
 
+def vectorize_delta(audio: np.ndarray) -> np.ndarray:
+    return add_deltas(vectorize(audio))
+
+
 def vectorize_inhibit(audio: np.ndarray) -> np.ndarray:
     """
     Returns an array of inputs generated from the
@@ -70,12 +84,15 @@ def vectorize_inhibit(audio: np.ndarray) -> np.ndarray:
     return np.array(inputs) if inputs else np.empty((0, pr.n_features, pr.feature_size))
 
 
-def load_vector(name: str, vectorizer: Callable = vectorize) -> np.ndarray:
+def load_vector(name: str, vectorizer: Callable = None) -> np.ndarray:
     """Loads and caches a vector input from a wav or npy file"""
     import os
+    vectorizer = vectorizer or (vectorize_delta if pr.use_delta else vectorize)
 
-    save_name = name if name.endswith('.npy') else os.path.join('.cache', str(abs(hash(pr))),
-                                                                vectorizer.__name__ + '.' + name + '.npy')
+    save_name = name if name.endswith('.npy') else os.path.join(
+        '.cache', hashlib.md5(
+            str(sorted(pr.__dict__.values())).encode()
+        ).hexdigest(), vectorizer.__name__ + '.' + name + '.npy')
 
     if os.path.isfile(save_name):
         return np.load(save_name)